Last week, at the onset of our Cybersecurity Month blog series, “The Barn Door is Open”, we argued that banning Artificial Intelligence (AI) outright is an illusion of control rather than actual defense. AI tools arrive through a browser, through Software as a Service (SaaS) vendor update, and embedded inside hardware you have already procured. Blocking web URLs does nothing to address the other two entry points.

The real challenge is architectural: Where can a model sit, what is it allowed to reach, and who remains accountable for what it produces?

In December 2025, the Cybersecurity and Infrastructure Security Agency (CISA) and the Australian Signals Directorate's Australian Cyber Security Centre (ASD's ACSC) published Principles for the Secure Integration of Artificial Intelligence in Operational Technology. These guidelines provide an actionable framework to rein in unmanaged AI use and learn how to co-exist with it.

Three Doors AI Uses to Enter

In our broader Cybersecurity Month blog series, we outlined the common pathways AI takes into a utility. Those entry points look like:

Direct Web Access (Shadow AI)

An employee opens a public generative AI chat tool in a browser and pastes in a paragraph of operational notes. The traffic travels over a standard web connection to a short list of known addresses. Secure web gateways and browser policies can log it or block this traffic relatively easily. This is the shadow AI most organizations focus on and is by far the easiest of the three to govern.

Embedded in Sanctioned SaaS

When an AI assistant lives inside approved business software (such as Microsoft 365), its traffic blends seamlessly into the routine authorized application sessions. There is no new address to block, and network perimeter inspection sees only standard traffic with a trusted vendor rather than the prompt payload inside it. Governing this door requires controls at the application layer: admin consoles and the audit logs.

Endpoint and Operational Technology (OT)

A model built into the operating system, running on an engineering workstation, or embedded in an OT device may never generate traffic that your network tools would notice. Nothing crosses the boundary you are watching. Oversight at this layer relies on asset management, configuration audits, and rigorous procurement vetting rather than network filtering.

The first door is the easiest to watch. It is often the only one that organizations measure. The problem is larger than that view shows. IBM's 2026 Cost of a Data Breach Report found that unapproved AI use played a part in 43% of the breached organizations it studied. The year before, the figure was 20%. If you only inspect web gateway logs, you are measuring a fraction of the exposure while mistaking it for the whole picture.

As one travels deeper through the network, visibility drops significantly further down into an industrial environment. In the enterprise and business layers, AI traffic and activity leave a clear trail in web proxies while passing through web gateways, DNS logs and SaaS audit trails, so one can have some visibility. At site operations and supervisory control, an AI model running locally leaves little trace. At the controller and field-device level there is often no agent to install and no log to read, so the sole visibility into AI activity on those devices is what the vendor explicitly discloses.

Figure 1 shows the three ways AI enters a utility, and how easy each one is to see.

Model Placement

Your organization already possesses the foundational structure to contain AI safely: segmented network zones and conduits, the Purdue Levels, and the established boundary inspection rules for traffic.

A model at Level 4 can still use plant data, for example data from the historian. The data comes up to the model. The model does not reach down for it. A replicated historian or a scheduled export in the demilitarized zone (DMZ) between Levels 3 and 4 gives the model a copy to read. The model never connects to Level 3 or below. It has no access to devices or their configuration.

An AI model in a conduit will travel between the supervisory control layer and the corporate layer. A replicated historian, a scheduled export, or a message broker in the demilitarized zone between Levels 3 and 4 hands data upward, allowing the model to read the copy as part of these processes. Although this may be useful, allowing an AI assistant to query or issue commands from the OT layer creates a high-risk conduit into the process that must be strictly authenticated, documented, secured, authorized and continuously monitored, like any other corporate application that reaches information in that layer.

Where the model runs is a critical decision that requires a clear view of your organization’s network: a vendor-hosted application sends prompts and context outside your network’s boundary, so the egress path and the vendor's data retention policies and training terms serve as your primary defense and become the security controls you will use to regulate usage.

An alternative is self-hosting (private cloud), which keeps the data inside your perimeter and shifts the burden of model hardening, vulnerability management, patching and high computational infrastructure costs onto your internal team.

Figure 2 maps the location where AI may sit in a utility and marks where these processes will flow across the DMZ.

Tool Access

The National Security Agency's Artificial Intelligence Security Center published an information sheet on Model Context Protocol security in May 2026. It names tools invoked on the fly, implicit trust between components and shared context as systemic risks the protocol does not inherently solve. Weak access controls, token lifecycle gaps and thin audit logging sit alongside them.

Every tool an agent can call is a path to data or to another system. Treat it like any other connection. Name it, give it one job, require it to log in, and log its use. The OWASP Top 10 for Large Language Model Applications ranks prompt injection as the top risk. Our AI Security Series explained this attack. A tool that returns text from an attacker is one way to deliver it.

Access control is where many organizations fall short. Of the organizations in IBM's 2026 study that had an AI-related breach, 92% did not have proper access controls for AI, such as role-based access and multifactor authentication.

Keeping It Reined In

To co-exist safely with AI across a utility, governance must be backed by verifiable telemetry. An approved tool catalog only works if network, identity, and host logs actively confirm what is running. Maintain your approved tool catalog on a quarterly review cycle, accelerating reviews whenever vendor terms change, models update, or integration scopes expand.

Map the telemetry to each door.

Each door AI uses leaves records of its use in a different place.

• On Direct Web Access: Evidence is in DNS query logs and web gateways. A lookup for a named AI service shows that a device reached it. The prompt itself only becomes visible where TLS inspection is switched on for that traffic; without it, the gateway records the name of the service and the amount of data sent, but not the prompt. Netskope's Cloud and Threat Report 2026 found that the average organization records 223 data policy violations a month that involve generative AI. Allow only the AI services that your organization has approved, usually one or two, and block all other AI services at the web proxy and in DNS. A log alone does not protect you. It shows that data left. It does not stop it. Turn on inspection for the approved services so that you can see what goes out. If you use Microsoft Purview, network or browser data loss prevention (DLP) can also record use of AI apps that your organization has not deployed. These records appear as AIAppInteraction entries in the audit log. You must turn on pay-as-you-go billing to get them.
• On embedded software: Evidence sits in the tenant. Microsoft Purview’s audit log for Copilot and AI applications records which user interacted with Copilot, when, and which files, sites or other resources it accessed to answer, under the CopilotInteraction record type. Non-Microsoft AI apps registered in the tenant are logged as ConnectedAIAppInteraction entries. In Entra ID, the delegated and application permissions granted to enterprise applications show which third-party AI tools have been consented into mailboxes and files. Microsoft notes that revoking a grant does not stop users from consenting again, so the review needs a user consent setting behind it.
• On Endpoint and OT: Evidence is in the endpoint inventory. Microsoft Defender Vulnerability Management builds its software inventory from the same signals Defender for Endpoint uses for detection and refreshes it every three to four hours. On Linux it lists only software installed through native package managers such as rpm, dnf and yum, so a model runtime unpacked by hand will not appear. Controllers and field devices that cannot run an agent stay outside this view entirely and will require vendor disclosure.

Tune the content detectors to your own material:

Asset inventories, relay settings, network ranges and one-line diagrams are what matter in a utility, so build your rule set around these types of confidential data.

Reining the horses in is not the same as putting them to work. The next part of our mini-series covers where to go from here.

References

Knowing where Artificial Intelligence already lives inside your organization is harder than it sounds. The first three parts of our AI Cyber Security Series set the ground: Part 1 covered what attackers are doing with AI, Part 2 built the policy, and Part 3 ran the five gates to help implement the policy. Through October, for Cyber Security Awareness Month, a shorter series looks at living with AI once it is already inside: reining it in, coexisting with it, and pointing it at work.

The horses are out of the barn and now we have to see where they went. The hardest part of governing AI inside a utility is not always writing the rules; it is also knowing every place the technology has already reached, and which of those places your existing tools can see.

The Many Ways AI Enters an Organization

A staff member opened a browser tab with an AI Chat, a vendor pushed an update that switched an AI assistant on inside software you had already approved, a device shipped with an AI model embedded. Each of these paths occur in organizations and require a different set of security controls.

 What has changed these recent years is the demand. The question used to be whether staff would use AI tools at all, now staff that wanted nothing to do with AI are willing to adopt this technology to streamline processes. This shift in appetite for AI tools is hard to switch off, as many staff now rely on AI to keep certain processes going, even with a human in the loop.

 Attackers have made the same move. Part 1 covered many ways where AI has been integrated into offensive security operations. Your staff are not the only ones who have found use of this technology.

AI Already Touches Your Operations

It is essential to see how far the reach of AI goes in your organization before you decide what to do about it. AI does not arrive with a single product, it is embedded into multiple areas such as a feature switched on inside tools your teams already use, and as capability your vendors already shipped. For a utility, that reach can run from the front office to the plant floor.

Consider where AI already sits in a typical operator:

• The Microsoft 365 drafts and summarizes. An assistant in the word processor or the email client will condense long email chains and operational details through the model to produce a result.
• The meeting platforms may transcribe and summarize engagements by default, including during sensitive topics such as outage bridge calls and vendor design reviews where relay settings, network segments and restoration steps are discussed openly.
• Engineering workstations running code assistants. Control engineers who ask a built-in assistant to help write or explain a script pass fragments of logic, this could include process details.
• Vendor Operational Technology (OT) products now carry AI of their own. Cybersecurity and Infrastructure Security Agency (CISA) noted in December 2025 that some OT devices ship with it built in, sometimes needing internet connectivity to work.

These applications of AI never asked permission to be in your system. These implementations are in place where your own information can now meet an AI model.

Banning AI Outright is a Weakness

Applying a ban on AI use on its own fails. A blocklist to AI sites closes only one way of entry, and new model endpoints appear faster than ever before. Embedded AI behind SaaS platforms sits inside tools you cannot block without breaking the business, and a staff member on a personal phone using AI cannot be monitored without infringing staff privacy. A ban just moves AI usage to a place where you cannot control its uses. On the OT network, blocking by default stays the rule, and should stay as such. Certain exceptions may arise and will be discussed further in the series.

The cost of having these blind spots could result in security incidents. In IBM's 2026 Cost of a Data Breach study, 43% of the organizations studied had a security incident involving unapproved AI use, up from 20% the year before. For a utility, the material that leaves this way is rarely trivial: asset inventories, network ranges and relay settings, the details an attacker would otherwise have to work to collect. Once it is typed into an unvetted model it is out of your control.

The Path Forward

The answer to preventing AI from scaling your wall is not to make the wall taller. It is to have a clear view of everything so that you may control it. Part 3 of our main series set out the five gates for that discovery and decision work.

We will expand on this notion by mapping out where your organization should AI Exist, coexisting with the AI you cannot switch off, and where a model may sit on your network.

The horses are already out and herding them back in is not the goal. Knowing every gap in the fence is.

References

Writing the policy is the easy part. IBM’s 2025 Cost of a Data Breach research, drawn from 600 breached organizations studied by the Ponemon Institute, found that 63% had no AI governance policy in place to manage AI or to prevent staff using it without approval. The Cybersecurity and Infrastructure Security Agency (CISA) and its partners published their principles for AI in operational technology in December 2025. Neither threat intelligence nor written policy alone can fully close the distance between documentation and practice.

This segment focuses on moving from policy to the implementation of your strategy, starting with the five gates.

Entering the Gates

Before we begin implementation, we need to know where we stand in the implementation path. Think of the five gates as a pre-flight checklist. The gates are linked to five sequential questions that any AI use case must clear before it is deployed anywhere in the organization.

We advise running the gates as a working session with the key stakeholders present in the room:

• Your security lead.
• Operations lead.
• Compliance lead.
• IT team.
• Procurement team.
• A senior manager to make on-the-spot decisions.

Every question should be answered as yes, partially or no. Where the answer is yes, the supporting evidence should be documented.

Run the gates once to establish a baseline of where the organization stands, then evaluate each new use case individually. Stop at the first gate that is not passed. Remember that answering no is not a failure; it is simply the first item on your roadmap.

Clearing all five gates shows that an organization is ready to run AI in its business layer. Most organizations will not clear every gate on the first pass; this is a normal outcome that highlights the starting point for your roadmap to successful implementation.

Gate 1. Defining the Need

1. What problem are we trying to solve?
2. Does something we already own, or a minor change to an existing process, solve the issue without adding a new system?
3. Is this request coming from an operator with an operational problem, or a vendor pushing a product?

The statement “We need an AI assistant” is not sufficient because it only defines a tool and not the business problem. By contrast, “Field crews spend too long finding the right procedure, and an AI assistant would make that process more efficient” is a more appropriate use case.

Question 2 forces a look at existing assets. CISA advises operators to consider solving the problem with an existing tool before introducing complexity.

Question 3 confirms that the request comes from a real operational bottleneck, not vendor marketing. A feature offered by a vendor is not the same as an issue raised by an operator.

Gate 2. Shadow AI Discovery

1. Which AI features are already switched on, across software, email, documents, meetings, and engineering tools?
2. Which vendors have injected AI into a product we already run via automated updates?
3. Are staff using personal accounts for work tasks, and do we know that for a fact or are we assuming it?

This gate is an internal discovery exercise. Admin consoles, license entitlements, vendor release notes, proxy logs and your own staff can all help answer these questions. IBM found that one in five of the organizations it studied had a breach linked to shadow AI. CISA notes that some Operational Technology (OT) devices now ship with AI built in, so this review will cover the plant floor as well as the corporate office. A written inventory listing each tool, its users, its data tier and its status will clear this gate.

Gate 3. Data Classification & Safety

1. Do we have a data classification scheme, and do people actually use it?
2. Have we identified our highest tier information and mapped it into that scheme?
3. Can we tell a member of staff in one clear statement what they are prohibited from typing into a chat window?

Evaluate your data classification tiers in the organization by handing a few documents to several staff members and asking them to classify them independently. If their answers differ, the scheme has not been implemented well enough for staff to apply it consistently. A good benchmark is CISA's list of engineering configuration data, which includes network diagrams, asset inventory, and schematics. Electric utilities have a head start under the North American Electric Reliability Corporation (NERC) Critical Infrastructure Protection (CIP) 011 standard; they already identify and protect information about their BES Cyber Systems, which gives them a tested process and a top tier to map that information into. The rest of the organization's data still must be classified.

Gate 4. Ownership & Total Cost

1. Who owns this program by name, rather than by committee?
2. Who reviews the output, and do they possess the subject-matter expertise to catch a confident, plausible mistake?
3. What is the true annual cost, including human review time, not just software licenses?

A committee can advise on ownership, but one named individual must take ultimate accountability for the program. CISA warns that AI can hallucinate, which means it can give a plausible but false answer, and should almost certainly not be used to make safety-critical decisions in OT environments.

Cost is where most implementation plans break down. Factor in licenses, integration, logging, training, and the hours staff spend verifying the integrity of AI output. This gate is passed when the program has a named owner, a qualified reviewer for each approved use, and a true annual cost that accounts for review and verification time.

Gate 5. Auditability & Incident Response

1. If a regulator asks whether AI touched a specific submission, can we answer from concrete records rather than memory?
2. Are accounts individual, and is usage logged?
3. Does our incident response plan cover sensitive material inadvertently fed into the wrong tool?

Regulated operators already live by audit-ready evidence, and AI use must meet the same standard. CISA recommends collecting flow and access logs for AI endpoints, tracking data leaving by asset and identity, and logging AI actions under an identity separate from any user or machine accounts. IBM found that 97% of breached organizations reporting an AI-related incident lacked proper access controls around the AI tools. Passing this gate requires all AI tool sessions to have single sign-on, reviewable audit logs, and an established incident response plan for tool misuse.

Figure 1 summarizes the five gates as a checklist, with the evidence that counts as a pass for each one.

Making Implementation a Daily Practice

Clearing the gates does not mean the work is finished; it means you are finally cleared to start implementation. The following steps are recommended to help successfully bring AI under control within your organization.  Each step represents a small but important part of the bigger picture and contributes to the successful implementation and adoption of AI tools.

Classify Your Data, and Verify Adoption

Part 2 established a four-tier classification scheme. By embedding these tiers directly into document templates and shared file structures, the label travels alongside the file. From there, organizations must verify that the staff are applying the labels correctly. In May 2025, the National Security Agency (NSA) and its partners released AI data security guidance advising organizations to classify input data, apply strict access controls to each tier, and assign AI-generated outputs the same classification level as the source data that fed them. Practical ways to audit label adoption include conducting blind classification tests with staff, scanning for unlabeled documents, and spot-checking AI outputs against input tiers.

Publish the Approved Tool List

Maintain an AI-approved tool list that pairs each tool with the highest data tier it is cleared to handle. Always opt for enterprise agreements that explicitly prohibit vendors from using your data to train their models. Because vendor terms may change without warning, every entry must carry a review date.  For example, on April 24, 2026, GitHub Copilot began using interaction data from its personal Free, Pro and Pro+ plans to train and improve models in 2026 unless the user opts out, while the business and enterprise plans were protected. Ensure that this list is incorporated into IT policies and procedures so that staff members always have a concise, single, and definitive reference that clearly indicates what is permitted at their workstations.

Train Staff on Today’s Signals

Fold AI awareness into your organization’s training program it likely already runs, like the NERC CIP-004 program for electric utilities. The old ways to recognize a malicious request are gone. Poor grammar no longer marks a phishing message, and a familiar voice or face on a call is no longer proof of who is speaking. Train personnel to recognize the signals that can still be detected: unusual requests, odd changes to access, urgency and secrecy, and verification through a channel the sender did not choose. In addition, reviewers need dedicated training on how to check AI output against its sources and record their processes on how to efficiently verify output integrity.

Run a Scoped Pilot

For an AI tool that has already passed through the gates and is ready for implementation, always run a test pilot for this tool with all security measures active. Keep the pilot strictly in the business layer using Public or Internal data; this ensures that performance is easy to measure and any errors are simple to reverse. Effective use cases for a pilot include summarizing published standards and filings or drafting non-sensitive internal procedures. Under no circumstances should the control path, BES Cyber System Information (BCSI), or any data feeding a safety decision be included in this pilot.

Figure 2 summarizes the four steps of daily practice, from classification through to this pilot, and the test that says each one is done.

Write a brief pilot plan outlining the core problem, the objectives, potential failure risks, team members, the data tier limits, success metrics, and termination triggers.

Restrict the trial to an approved tool, mandate that a qualified reviewer check every output, and track the exact hours spent on review, which shows the true cost of the tool in both staff time and financial capital.

Following the pilot, evaluate your findings to make a clear decision: Expand, adjust, or stop. Shutting down a pilot that did not earn back its review time is not a failure but a sign that the governance program is working and saves money overall. Conclude the pilot with a declared result, a final decision, and the updated policy reflecting your lessons learned.

Figure 3 below walks through the pilot, from setting the boundaries to closing it out.

Implementation Is Not the Finish Line

The horses are already out of the barn, and trying to herd them back inside is no longer an option. The real challenge now is reining them in, getting them under control, and figuring out where they can actually pull their weight in your organization.

Once AI is properly reined in, the next objective is finding the precise tasks where it genuinely earns its place within your organization.

References

Procuring an AI tool is the easiest step in this process. The selected tool will not be able to decide what data your staff may put into it, who reviews the output, or who is accountable when things go wrong. So, the question is what do you need to do when AI is introduced within your technology stack and placed directly into the hands of your own staff.

For many utilities, this reality has already arrived:

• Staff have personal accounts on AI tools that are not approved.
• Meeting platforms transcribe and summarize by default using AI-enabled tools.
• Vendors have added AI features to software you already run, delivered in updates you already applied.

In several organizations, the decision to adopt AI was never actually made. Staff took matters into their own hands. The question is not whether to allow AI, but whether the use already underway is governed, and that requires a policy.

This second part of our AI Cyber Security Series focuses on what it takes to govern AI adoption from within.

The Requirement for an AI Policy

With the rapid advancement and adoption of these new technologies, organizations have not been able to completely govern acceptable usage, safe practices, or incident response with regard to artificial intelligence.

Users have begun using these tools without proper vetting or risk assessment. That practice is known as Shadow AI. IBM defines it as the following.

“the unsanctioned use of any artificial intelligence (AI) tool or application by employees or end users without the formal approval or oversight of the information technology (IT) department.”

The instinct in many IT departments is to ban it. But users have gained real productivity from these tools, taking them away can be counterproductive to organizations. Somebody who finds a tool useful will carry on using it on a personal device, and the confidential work then leaves your environment with no record of what went along with it. A policy that says yes with boundaries gives you a foundation for navigating the threat landscape.

AI-embedded systems already run inside your organization. Assistants in the office suite, meeting transcription tools, browser add-ons and code assistants in engineering environments are present in systems used daily by every type of employee. Most of them have no “disable” option.

The Cyber security and Infrastructure Security Agency (CISA) and its partners note in their December 2025 guidance that some Operational Technology (OT) devices now ship with built-in AI, in some cases needing internet connectivity to function. Whether your organization uses AI is no longer a question the IT department can answer on its own.

Without written governance, you have no boundary on what data leaves the organization, no record of where that data was used, and no answer when a regulator or a client asks whether AI touched sensitive information. Part of the threat landscape is how your own internal material becomes training material for somebody else's AI model. Anything typed into an unvetted model is out of your control from that moment, and for a utility, that material can include asset inventories, network details and more.

Policy Scope

A policy covers everyone, with no exemption for seniority, and every tool, including the AI features inside software you already own. Contractors working on your systems sit inside the scope as well.

Data Classifications

An AI policy requires a data classification scheme behind it. The moment a member of staff is about to paste something into a chat window, they need a clear rule on what is permitted and what is prohibited. This may result in unknown information disclosures within the organization.

The four tiers in the figure below work for most utilities:

What Data Classification Does Not Enforce

A classification scheme tells people and tools what matters. By itself it stops nothing, and that is where AI policies most often fail in practice. Once a member of staff selects a paragraph and copies it, there is no file and no metadata left, only text, and that text can be pasted anywhere. To combat these technical gaps, we recommend the following best practices:

• Put enforcement where the traffic passes. An endpoint agent, a managed browser or a web gateway can inspect a paste into an AI site. 
• Write detectors for what the top tiers of your classification actually look like in your organization, such as asset tags, relay settings or address ranges. 
• Log what cannot be blocked. What people are actually pasting is the input to the next revision of the tool list. 

Approved Tools

Classification applies to tools as well as to information. The policy should name which tools are allowed and what class of data each one is cleared for. Evaluate each tool in use. Questions regarding the assessment may include:

• Where is data processed and stored?
• Does the tool train on customer input?
• Does the tool support single sign-on and audit logging?
• Does the tool conflict with requirements you already work under?

Terms of service can change quickly. Since 24 April 2026, GitHub Copilot has used interaction data from the personal Free, Pro and Pro+ plans to train and improve its models unless the user opts out. The Copilot Business and Copilot Enterprise plans are not affected by that change.

Changes like this will continue to come, so set a cadence for reviewing the tool inventory.

Acceptable Use

Guidelines for what is not allowed should take a clear position on what users are already doing. It is recommended to apply the following best practices:

• No personal accounts for work.
• State the tiers from your organization’s data classification scheme that may not go into any AI tool.
• No AI output is treated as fact until a qualified member of staff has reviewed it.

Accountability

AI output is raw material. It is not finished work, and a qualified individual should review it before any of it is accepted. AI can never be held accountable. It is a tool that acts on a user’s instructions, so the user is always responsible. An IBM training manual from 1979 captured the same idea. 

A computer can never be held accountable; therefore, a computer must never make a management decision.

An individual who submits work produced with AI owns it. To enforce that accountability, user accounts should be individual rather than shared, which keeps logging and auditing accurate.

Transparency

Transparency should be encouraged in any organization that uses this technology. AI is both an exciting and a divisive talking point depending on who is asked.

Organizations and individuals who get called out for using AI are usually the ones who did not declare it. So, the use and its purpose are best disclosed in advance. Consider how AI was used, and whether to disclose it:

• If AI helped with research, structure or formatting and the author checked everything, no disclosure is needed.
• If AI drafted substantive content used in the final work, it is advised to say so.
• If AI generated analysis, summaries, or interpretations that shaped the conclusions, disclose which sections relied on AI output and describe how it was reviewed or verified.
• If AI tools processed sensitive, proprietary, or personal data during the work, disclosure should note this alongside the data-handling safeguards used.

Policy Limits

Some things an AI policy will not reach, and they need separate attention. A policy governs how your personnel use AI. It does not govern where AI sits in your network, what your vendors have embedded in your equipment, or what you do when a model strays from the intended instruction set.

Network Segregation

Language models belong in the business layers, working on data taken out of systems. Nothing should reach into the control path on its own authority.

Push data out, and do not let anything reach in. CISA recommends push-based or brokered patterns that move summaries out of the OT network without granting persistent inbound access. That keeps the OT network segregated from the network where AI tools are configured.

Remember that AI is meant to make parts of the job easier, not to do the job for you. Keep operators practised enough to run processes this way and always keep a human in the loop.

Figure 2 maps this across the plant. At the bottom are field devices, then local controllers such as programmable logic controllers (PLCs) and remote terminal units (RTUs). Above them sit the supervisory systems operators work with, including supervisory control and data acquisition (SCADA), distributed control systems (DCS) and human-machine interfaces (HMI). Language models belong at the top, in the enterprise and business systems. 

Vendor AI

What AI your vendors have embedded is a question that is usually left unanswered. Vendors are adding AI to OT products, sometimes with a connectivity requirement attached. AI capabilities in these systems include:

• Predictive maintenance
• Asset discovery
• Automated incident response

When engaging with an OT vendor for a new system, ask these questions as groundwork for understanding what AI the system carries:

• What AI is in this product?
• Where does the model run?
• Does the model train on our operational data?
• Can it work without continuous access to your cloud?
• Can we turn the feature off and keep it off?

Incident Response

Incident response plans are usually drafted once and do not go through major revision until they are executed during an incident. As seen in part 1 of our series, attackers of all skill levels are equipping themselves with this technology. Plans should cover AI related incidents including the following:

• An attack against or through an AI system can include manipulated input, poisoned data, or a model acting on instructions hidden in something it reads.
• AI system failure, where a model drifts until its outputs no longer match the plant.
• Self-inflicted data disclosure, which is sensitive material typed into the wrong tool.

CISA's guidance tells operators to update response plans for malicious activity directed at an AI system and for AI system failure. Add a section covering what the process does when an AI-assisted system is unavailable, then run it as a tabletop exercise.

Standards

The North American Electric Reliability Corporation Critical Infrastructure Protection (NERC CIP) standards already govern Bulk Electric System Cyber System Information (BCSI) handling, and AI use should not create an exemption. On the voluntary side, the National Institute of Standards and Technology (NIST) AI Risk Management Framework and its Generative AI Profile are good starting points.

Sector-specific guidance is still being written. NIST released a concept note in April 2026 for an AI RMF profile on trustworthy AI in critical infrastructure, and AI RMF 1.0 is itself under revision.

The Five Gates to Assess AI Usage

Before any action, it is worth assessing where your organization stands on AI use today. One way to do that is to run every proposed tool, and every tool already running, through five gates: do we need it, do we know what is already running, can we feed it safely, can we carry it, and can we prove it. A tool that fails a gate does not proceed until the gap behind that gate is closed. Organizations that skip this step are the ones most likely to get hurt.

From Policy to Practice

The policy is not the finish line, it's actually what lets you start. Most utilities that work through the five gates find the first gap is not technology at all, it is how people understand data and how they handle it, and both are within reach when the strategy is set up properly.

Our next part in the series works through each gate in detail and more. 

References

Artificial Intelligence (AI) has not given attackers new capabilities. It has made the ones they already had cheaper, faster, and available to people who could not run them before. Here is what that changes for the utilities, water systems, pipelines and transit agencies we work with.

In November 2025, Anthropic reported disrupting a cyber-espionage campaign, attributed to a nation-state actor, in which an AI model carried out an estimated 80 to 90 percent of the tactical work across roughly 30 target organizations. People still chose the targets and made the key calls.

The United Kingdom's National Cyber Security Centre (NCSC) had already published the finding, in January 2024 and again in May 2025: AI makes the capabilities attackers already had more effective and more efficient and puts them within reach of more people. Check Point’s Annual AI Security Report 2026 put it in one line in July 2026: "AI is now operating attacks, not just enabling them".

The path from research to worldwide usage took about a decade. The timeline below marks key points in the evolution of this technology that continues to reshape our world.

What Attackers Are Doing Now

Social Engineering Without the Tells

The bad grammar that gave phishing away is gone, and the model does background research too. A generative model is now able to write complete pieces in fluent business English, even able to be personalized to suit various writing styles that can be indistinguishable from a human. Open-Source Intelligence (OSINT) automation tools such as SpiderFoot and theHarvester allow for rapid intelligence gathering of entire organizations and their members.

In 2024, staff at the Hong Kong office of the engineering firm Arup transferred funds after a video call in which every participant except the victim was synthetic. Arup confirmed it was targeted with fake voices and images; with the reported loss being about 25.6 million US dollars according to police figures.

The Federal Bureau of Investigation (FBI) issued a public service announcement in May 2025 dealing specifically with voice cloning of senior US officials. Furthermore, Furthermore, Microsoft has documented a Chromium extension in June 2026 that was passing itself off as a legitimate AI assistant, impersonating a real AI vendor with look-alike branding and a typosquatted address. This extension was hijacking browser searches and capturing keystrokes.

Vulnerability Research Advancements

Google's Threat Intelligence Group reported in May 2026 that threat actors are feeding public vulnerability data into AI tools and documented the first zero-day exploit believed to be developed with AI assistance. A zero-day is a flaw with no fix available when it is first used.

Defenders get the same lift from the same technology. Google's Big Sleep agent found its first real-world flaw in widely used software in November 2024, later fixed by developers after they were notified. Recently, Microsoft has announced that for July 2026 patch Tuesday, a record 622 CVEs have been disclosed, Microsoft and analysts have suggested AI tooling is behind the rising volume.

This is certain to bring an increase in security patching volume it is possible that the regular update schedule will no longer be enough.

Criminal Models for Rent

Commercially available models contain guardrails and external safety controls to prevent malicious use according to usage guidelines. However, attackers have found ways to circumvent these controls and use commercial models to write malicious code and create specialized payloads.

New ways through the guardrails keep appearing, some startlingly simple; research published in November 2025 found that rewriting a forbidden request as a poem defeated safety controls across a range of models

Currently, an emergence of malicious LLMs have been available from dark web vendor sites. Palo Alto Networks Unit 42 catalogued the current generation of malicious LLMs in November 2025, including WormGPT 4 and KawaiiGPT. These are wrapped or jailbroken commodity models where researchers could not tell whether these are genuinely retrained models or ordinary ones held jailbroken.

Malware That Asks a Model For Help

Google's Threat Intelligence Group documented the first malware families that call an LLM at runtime in November 2025. One of them, PROMPTSTEAL, was observed in live operations run by a state-sponsored group. The other, PROMPTFLUX, interacts with its Gemini artificial intelligence (AI) model Application Programming Interface (API) to write its own source code for improved obfuscation and evasion. PROMPTFLUX is as of now experimental, Google states it cannot compromise a network or device, and its self-rewriting function was switched off in the sample analyzed. However, it sets a concerning precedent for things to come.

When Your AI Becomes the Target

When deploying AI systems in your organization. These systems can possibly act as attack surfaces through the employment of supply chain style attacks.

An AI assistant that reaches across mail, files, chat and calendars is useful for its reach but is also exposed by it. Text sitting in a document or an email can be written to read as an instruction, and the assistant may follow it.  This is known as Prompt Injection; and today it is still mostly researcher work.

Prompt injection vulnerabilities are now being disclosed in AI-enabled applications. A Microsoft 365 Copilot flaw disclosed in June 2025 could extract data with no user action, and Microsoft fixed it and reported no evidence it was ever used.

This pattern reaches past any single vendor. The National Security Agency (NSA) published a cyber security information sheet on Model Context Protocol security in 2026, stating that the layer where AI agents talk to tools is now attack surface. 

What This Means for Your Sector

Exploitation Is Outrunning Remediation

The gap between a flaw becoming public and a working attack existing is closing. VulnCheck found that 28.96% of newly exploited CVEs in 2025 were hit on or before the day they were published, and Axios reported on 21 July that research team watched an attacker take just nine hours to create a working exploit for a critical flaw after it was disclosed and then point that exploit at government customers.

The Skill Barrier Is Dropping Fast

CrowdStrike's 2026 Global Threat Report describes AI as lifting less sophisticated actors while amplifying the most advanced, with an 89% year-over-year rise in AI-enabled adversary activity. This will be a prevalent threat as AI models improve their capabilities, and get incorporated into different sectors, amplifying the attack surface.

Your Own Transparency Feeds Their Research

All areas where an AI agent can navigate to are at risk of becoming its reading material. This includes asset inventories, procurement records, standards submissions, interconnection filings, conference talks. Google's May 2026 findings showed actors feeding CVE data and public vulnerability archives straight into AI tools. Any published detail about which controllers and firmware your organization runs allows an attacker to build a profile for faster vulnerability research.

AI at the OT Boundary

Dragos documented a case in May 2026, where an unattributed actor used commercial AI tools during the compromise of a municipal water and drainage utility serving the Monterrey area of Mexico. After reaching the corporate network, the AI found a gateway into the industrial control environment, worked out on its own that it mattered, researched the vendor, built a password list and ran an automated guessing attack. It failed, and no control systems were breached. Current AI models give attackers no new industrial capabilities, and what they change is speed. The attacker had no industrial background. The AI supplied it and found control-adjacent systems within hours.

Then, in late July, on 26 and 27 July 2026, more than thirty Minnesota community water systems were hit in a coordinated attack; in one town the treatment plant controls went down, and the water tower could not be filled for over an hour. Within days the FBI and the Environmental Protection Agency warned that utilities in at least seven states had reported incidents, with effects including loss of pressure and flooding. Attackers reached control devices left exposed on the internet and set passwords on them, locking operators out of their own equipment.

Although there has not been a confirmed OT related attack using AI, it is getting clearer with each attempt that it is only a matter of time.

The Bottom Line

Phishing worked before AI. Weak remote access worked before AI. AI made current capabilities both cheaper per attempt and better per attempt.

Federal guidance is coming but is not here. America's AI Action Plan directed by the Department of Homeland Security (DHS) in July 2025 to establish an AI Information Sharing and Analysis Center, the AI counterpart to the Electricity Information Sharing and Analysis Center (E-ISAC) and Multi-State Information Sharing and Analysis Center (MS-ISAC) channels your teams already use. It has not launched. A Cyber Security and Infrastructure Security Agency (CISA) official described it in February 2026 as still in development.

It is advised to consider the following when dealing with the evolving threat landscape of AI tools.

• Get system control devices off the public internet and put remote access behind a gateway you control.
• Train your organization to familiarize themselves with what attacks look like now. The signs on how to tell previously are no longer valid.
• Shorten the patch window on any internet-facing asset, the time-to-exploit will continue to get smaller.
• Reduce your AI attack surface by limiting what is publicly available from your organization, and put any AI tool you deploy through your existing third-party review
• Create an AI usage policy to better control and govern AI usage in your organization.

References

In our previous blog post discussing the Energy Transition Roadmap for Electrical Utilities, we looked at challenges and solutions for achieving Energy Transition objectives for the fundamental aspects of power systems – namely power generation, transmission and distribution, and customer demand. Today, we shift our focus to the technology needed to monitor and control electrical resources on the grid, and how they need to evolve in the Energy Transition journey.

There are many aspects of this Technology dimension:

• Power Production is monitored by an Energy Management System (EMS)
• Transmission Grid is monitored by a Transmission Management System (TMS)
• Distribution Grid is monitored by a Distribution Management System (DMS), and more extensively these days by an Advanced Distribution Management System (ADMS).
• Distributed Energy Resources (DERs) may be monitored by Distributed Energy Resource Management Systems (DERMS).

Providing Context - Stability through Inertia

For many decades the stability of the power system depended on the inertia of the large generators. Using reverse frequency droop settings on the governors of the generating units allowed for countering the results of faults on the system. If there was a loss of a large generator, it meant that there was more load than generation. This would cause the frequency to start reducing. The reduction in frequency would automatically cause the online generating units to produce more power to counteract the drop in frequency. This nearly immediate automatic response was then backed up by other system responses such as commanding the setpoints of other generators to increase to counteract the drop in generation. This resulted in the system remaining at a stable frequency and becoming rebalanced. On the other hand, if a large load was lost, the frequency would rise, the amount of generation would be automatically reduced, and the system would respond by lowering other generating unit setpoints.

An underlying principle of this methodology was the substantial inertia of the numerous large generating unit rotors that tended to keep spinning as required. This worked for many decades until entities began adding renewable generation, such as wind farms and solar farms - power resources that did not come with large inertia. The renewable generation resources may have had frequency control in the invertors connecting them to the grid, but in the early days of renewable generation operation, it was standard operating procedure to command them to disconnect from the system during faults. However, the loss of the renewable generation could worsen the frequency problem or make it better, depending on the fault type. If the fault was a loss of generation, then dumping more generation would make the problem worse. If the fault was a loss of load, dumping the renewable generation would make the system better.

Defining the Problem

The Renewable Energy Challenge - As the penetration of renewable generation is growing, power system engineers are modifying their approach to operating these resources. There is ongoing research and testing, and Grid Forming Inverters are being proposed as a solution. Grid forming inverters enable renewable generation resources to partly act as inertia-based resources. This means that a particular renewable generation resource can reduce generation in the event of a loss of load. However, in order for renewable generation to increase the output in response to loss of generation elsewhere, it must have some kind of energy storage that is rapidly available, such as a Battery Energy Storage Systems (BESS). A renewable resource with BESS that has enough energy stored can provide a reverse frequency droop characteristic in response to a fault on the system. Research is still needed on how the rapid frequency change is handled by the power electronics of the solar inverters or the wind farm convertors.

The key takeaway from this discussion is that as the system shifts to a large mix of renewable energy resources, it must also be built with tools that allow the power system to remain stable in response to faults.

Fault Current Detection - The other side of the question of inertia is related to faults on the system that can result in short circuit currents. Traditional rotor-based generators provide high short circuit currents into a fault because of their inherent large inertia. Renewable Energy Resources do not provide high short circuit currents in the same way, because they are connected to the grid through electronic based inverters. The lack of high short circuit currents in the event of a fault has detrimental effects on fault coordination, which is based on protection relays reading the high short circuit currents. While research is still ongoing in solving this challenge through novel protection schemes and relays, BESS could once again be a potential solution. For example, if a large solar farm also has BESS, this could enable it to shift the load and provide additional current into a fault to enable the protection systems to work.

Voltage Support - The other aspect of power systems is voltage support. The system needs reactive power to ensure that the voltage profiles of the system are within acceptable limits and that the power will flow from the resources to the loads. This means that large renewable resources have to be planned with reactive resources as required. Many Wind Farms and some Solar Farms have been built with additional Reactive and Capacitive resources. The amount needed will be dependent on the existing voltage situation in the Grid. Voltage support may require the addition of reactive resources on site at the Wind/Solar resources.

Planning & Load Balancing - The historical procedure for day-to-day planning is to perform load forecasts based on weather and factor in any planned outages. This ensures there is power available to supply the load as it varies during the day according to normally expected load curves, and when affected by weather and unusual events. A given self-contained area will ensure that it balances its load and generation throughout the day as well as its purchase/sales of power with other entities. The area interties are planned with neighbours, and these planned flows are compared with actual flows on the interties to adjust internal generation. The difference between planned interchange and actual interchange is called the Area Control Error (ACE). The North American Electric Reliability Corporation (NERC) rules require each area to keep ACE near zero and take action to return it to zero when it diverges.

To summarize, the work of monitoring, analyzing and controlling a power system network involves the following main issues:

• Generation load balancing for frequency stability
• ACE management for additional generation-load balancing
• Reactive power placement and availability for voltage stability
• Use of Generation inertia and relay protection to respond to system faults
• Ensuring that sufficient generation is available for rapid deployment for response to contingencies but also scheduled and available to follow load profiles
• Scheduling and directing generation changes to follow the load curve
• Analyzing and preparing to react to new contingencies when the system experiences a contingency and may be in an unusual state 
  

The Bulk Electrical System and Energy Transition

Renewable generation is intermittent and, while it can be predicted somewhat accurately, a bulk electrical system with high penetration of renewable generation must always have backup generation ready to replace large amounts that may be lost. It will be necessary to have bulk-level BESS or similar storage available to mitigate the effects of rapid loss or change in output levels of renewable generation. The loss of renewable generation may also mean that the replacement generation is located in different areas, changing the pattern of flow within the power system.

System Response and Contingencies

The main issue with the transmission grid is that multiple contingencies from strong storms can make system response very complex or result in a large reduction in the BESS energy available. Hydrogen is being proposed as the Energy Transition replacement for natural gas-fired generation. This requires an excess of renewable generation to manufacture enough hydrogen for use during minimal availability of renewables and limited storage.

Monitoring, Analysis, and Control on the Generation Level

It will be important to have extensive weather monitoring along with AI pattern analysis. Utilities can rely on weather forecasters to plan from 7 days to 1 or 2 days ahead, but they will need large arrays of measurement points utilized by all utilities to understand upcoming conditions for more immediate planning.

Weather Monitoring and AI Integration

Large solar/wind monitoring arrays within a utility’s footprint will allow the utility to understand local wind and cloud cover and how it is moving, enabling more accurate predictions of renewable outputs and localized temperature changes. AI will be trained to provide predictions that can be compared with actual measurements for continuous improvement. Utilities can optimize their other generation alongside more accurate forecasting of intermittent generation to better use their storage, allowing for smoother generation changes and more efficient hydrogen generation planning.

Data Collection and Simulation

Data collected from monitoring can be tested in simulations to refine a utility’s understanding of operations concerning the sun, clouds, wind, and storage, improving efficiency over time. Utilities will also be able to use simulations of faults in different places to learn how to respond effectively to different combinations of generation, storage, and load, aiding in arranging reactive power resources to support voltage as power flow patterns change.

Existing and Needed Tools

Currently, utilities use numerous tools for monitoring, analyzing, and simulating to effectively plan, operate, and respond to contingencies. These include power flow, state estimation, contingency analysis, reliability assessment, simulations, and transient stability analysis. Utilities use tools to optimize the use of generation based on cost and fuel availability, and other factors and tools for optimizing reactive resources to maintain voltage profiles.

Flexible AC Transmission System (FACTS)

FACTS devices assist with controlling reactive flow. Utilities will continue to need these tools and substation monitoring while using additional monitoring of transmission lines to enable real-time thermal ratings.

Additional Requirements

Utilities will need to add large arrays of weather monitoring for wind and cloud cover. AI tools with large capacities are required to handle the vast amount of real-time information about actual wind/solar conditions and predictions. This will enable utilities to forecast intermittent resource production and optimize storage use more precisely.

Simulation Tools for Optimization

Since weather patterns vary, utilities may also need to simulate how wind and sun resources vary based on historical data to improve storage and hydrogen production optimization. Continued use of simulation tools will help understand different faults and system responses, with research on Grid Forming Inverters to ensure they respond effectively to contingencies.

Inverter Based Resources (IBRs)

Research indicates that IBRs may respond to system issues differently depending on the problem. System monitoring (e.g. use of field Phasor Measurement Units (PMUs) with very fast communications and central processing) must rapidly understand and inform IBRs of the exact problem (e.g., a single-phase fault on Phase A) to enable precise responses (e.g., injecting extra current on Phase A to interrupt the fault). This must occur within milliseconds, utilizing the smart software of IBRs for effective system support.

Distribution System and Energy Transition

The Energy Transition will significantly impact the distribution system. The presence of substantial renewable/intermittent resources on the distribution network may result in two-way power flows. Currently, power flows from generation through the transmission system and into the distribution system on a one-way trip to customers. Adding solar/wind to the distribution system along with BESS and other storage types will radically change power flows, potentially opening an energy market at the distribution level and introducing Distribution System Operators (DSOs).

Fault Response on the Distribution System

Currently, renewable power generation disconnects from the distribution feeder during faults, typically in response to zero voltage. This means renewable power is lost unless local BESS charges up. Traditional distribution utility responses to faults involve opening breakers feeding the fault. Many utilities now use feeder reconfiguration with automated/controllable switches, identifying fault locations, opening switches to isolate the faulted section, and re-energizing unaffected segments.

Advanced Distribution Management Systems (ADMS)

Utilities use networked meter reading for hourly meter readings and last gasp detection to identify power loss or faults. Smart meters providing additional interval data within the hour would help utilities analyze and understand customer loads, facilitated by AI tools marketed by various companies.

Outage Management Systems (OMS)

Most utilities have an OMS that utilizes data from controllable switches and meters to reconfigure systems automatically or recommend switching orders to return to normal configurations. The OMS often works in conjunction with an ADMS, which uses SCADA systems for extensive data management, incorporating data from Distributed Energy Resources (DERs) and providing it to Distributed Energy Resource Management (DERMS) tools.

Virtual Power Plants (VPP)

Many entities have created VPPs, conglomerates of numerous customers creating virtual resources. These include customers allowing load control during peak times, grouping small solar or wind resources with BESS for extra utility. Combined with renewable resources, BESS, and controllable load, VPPs form versatile resources, requiring extensive communication points and data management for effective monitoring and control.

Future of the Distribution System

The distribution system of the future may exhibit varying load curves daily due to sun, wind, temperature, and market conditions. Feeder protection systems must adapt to manage resources larger than loads, with power flowing back to substations. Utilities might find it beneficial to have large BESS at transmission substations to act as buffers, smoothing out load shifts and easing transmission system management.

Comprehensive Monitoring Needs

While the transmission systems will continue to monitor substations, capacitor banks, and FACTS devices, adding BESS and advanced monitoring to handle unpredictable load curves, distribution systems must monitor, understand, and analyze a larger number of status and analog points from numerous sites, controlling more devices and issuing requests to numerous devices.

Market and Communication Integration

A market will mediate the costs of resources and loads participating in it, while working with regular loads not engaged in the market. Effective operation will require extensive communication points, high-speed data processing, and secure, corruption-free data exchange.

The Grid Edge

• Monitoring transmission and distribution facilities with intelligent measuring devices (e.g., detailed transformer monitoring, advanced line monitoring, real-time monitoring of important load/gen/BESS customers).
• Physical control: switchable line devices (breakers, smart grid switches), controllable load/gen/BESS sites.
• Utility requests: asking customers to adjust their load/gen/BESS, general requests for emergency assistance, potentially communicating directly with customers' smart controllers.
• Price Signals: sending customers distribution-level load/gen/BESS price signals.
• Transmission level: continued generation/substation/transmission monitoring with additional monitoring (e.g., using PMUs for power angle monitoring) to enable accurate instructions to Grid Forming inverters.

Control Centers:

• Increased communication connectivity to the system.
• Interaction with power markets and their price signals to predict power flow trends and load centers with more complex load profiles, learning from weather impacts on them.
• Intelligent power forecasts for wind/solar to manage changes in storage use and hydrogen generation.
• Monitoring IBR-based resources to determine fault current adequacy settings.
• Using existing power system monitoring tools like state estimation, Volt/Var system balancing, and contingency tools that account for Grid Forming inverters' responses to different system issues, and synchronous condensers.
• Rapidly responding to system contingencies with instructions to resources and loads within milliseconds.

Communications:

• Utilizing various telecom technologies for rapid data interchange, ensuring real-time data availability to all entities.
• Exchanging historical data like load, use, weather, and solar/wind patterns.

Overall, achieving effective Energy Transition goals requires a coordinated effort across all facets of the transmission and distribution systems, supported by advanced monitoring, AI integration, and robust communication infrastructure.

Acumen has a team of professionals dedicated to helping utilities create a path to enable the Energy Transition. Book a call today with experts from our Operations Technology team to better deal with the technological challenges of this once in a generation energy transition!

Critical infrastructure continues to emerge as a prime target for various types of cybersecurity breaches.

Between June and December, 2025, the widely used text editor Notepad++ became the focal point of a sophisticated supply chain attack ^[1][2]. Through communication with Notepad++'s previous hosting provider as well as help from security researchers, several network-based indicators have been discovered ^[3]. This was not a direct breach of the Notepad++ source code, but rather a compromise at the infrastructure level of its former shared hosting provider ^[4].

Investigative efforts by external experts and security firms like Kaspersky and Rapid7 have attributed the campaign with moderate confidence to state-sponsored threat actors. The campaign was highly selective, focusing on high-value organizations and specific individuals in regions including Vietnam, the Philippines, El Salvador, and Australia ^[5].

Understanding The Attack Vectors

The primary attack vector begins with the WinGup auto-updater component, allowing access to the hosting provider's environment. Attackers gained access to the shared hosting server and were able to intercept and redirect update traffic intended for the official Notepad ++ site. Even after regaining server control, the malicious actors maintained internal service credentials until December, allowing them to continue directing targeted users toward malicious update manifests ^[6]. These manifests served compromised versions of update.exe, which were often NSIS installers designed to deploy further malicious payloads.

The result of exploiting this initial attack vector allowed the attackers to utilize multiple infection chains that evolved over several months to maintain stealth.

One method involved DLL sideloading, where a renamed legitimate binary, such as the Bitdefender Submission Wizard, was used to load a malicious library named log.dll. This chain eventually executed a custom, feature-rich backdoor known as "Chrysalis" which provided the attackers with extensive control over infected systems ^[5:1].

A decomplication of malicious file alien.ini used to download malicious .dll files, provided by Kaspersky cybersecurity.

Other chains delivered Cobalt Strike Beacons via Metasploit downloaders, often hiding their activity by abusing old vulnerabilities in legitimate software or using undocumented system calls like Microsoft Warbird.

Why OT and Engineering Teams Should Care

Notepad++ is a common tool found in critical infrastructure, used for workstations in engineering, operations, maintenance, and jump host machines to reach OT networks.

A compromise on these systems is not merely a "desktop issue", these devices often host:

• Privileged Access to SCADA servers and SCADA historians
• Direct Connectivity to Programable logic Controllers, RTUs, IEDs, and plant networks
• Trusted status for critical change events and windows

A common misconception is that OT environments are not at risk due to their isolation from external traffic such as the internet. However, this incident highlights that actions such as manual software transfers are vulnerable.

In a typical workflow, an engineer downloads software or updates in the corporate IT environment, copies them to "approved" media or a staging share, and then physically brings them into the OT zone to run on an engineering workstation.

If the installer was poisoned on the IT side during the initial download, OT segmentation does not undo that risk. Instead, this process effectively carries malicious code straight through your defenses and executes it inside your most trusted zone.

Exploiting this weakness turned a simple text-editor into a sophisticated supply chain attack affecting various organizations around the world.

Indications of Compromise

Malicious Traffic Indicators

A simple starting point for your Security Operations Centre:

Please note that these Indicators of compromise may be outdated as this is a developing investigation. More technical information and IoCs can be found through Kaspersky Cybersecurity and Rapid7 Labs investigations.

The Consistent Looming Threat of Supply Chain Attacks

These events highlight the severe risks inherent in supply chain attacks, a more frequently occurring type of security incident. These types of attacks weaponize the trust between a software vendor and their various partners. These could be third-party components, service providers, or even vendor clients. By compromising the distribution infrastructure rather than the application code itself, attackers can bypass traditional security perimeters and deliver malware to targeted high-value systems under the guise of legitimate software updates.

The highly selective nature of this attack by targeting only a dozen machines out of millions of users makes it exceptionally difficult to detect. Attackers are capable of rapidly adapting their tools and infection chains to evade signature based detection. This level of persistence and stealth allows state-sponsored actors to maintain a foothold in critical sectors like government, finance, and telecommunications.

Course Correction: Notepad++'s Response

This security incident underscores that open communication and transparency between a software vendor and its third-party providers are essential for managing supply chain risks. In this case, Notepad++ actively facilitated direct communication between the third-party hosting provider and the external incident response (IR) team. Collaboration allows security experts to have the necessary access to investigate indications of compromise, and required mitigations.

In response to the discovery of this incident, Notepad++ took immediate steps to secure the project's infrastructure. The entirety of Notepad++ has been migrated to an entirely new service provider, which is stated by Notepad++ to have significantly stronger security practices.

To prevent future hijacking of the update process, the WinGup updater was enhanced in the most current release (8.8.9). Furthermore, starting with version v8.9.2, the update server began returning signed XML (XMLDSig) to ensure the integrity of the update manifests ^[7].

The release of Notepad++ v8.9.1 served as a critical security update, incorporating these enhancements to block the specific redirection tactics used by the attackers. This version also introduced a new option during the installation process that allows users to disable the auto-updater entirely if they prefer manual management ^[8].

msiexec /i npp.8.9.1.Installer.x64.msi NOUPDATER=1

Recommendations and Preventative Measures

As supply chain attacks become more frequent and harder to detect, the trust between software vendors and critical infrastructure must be stronger than it ever was before. Proactive defense is essential for maintaining operational resilience.

If your organization needs assistance with implementing the remediation steps discussed through supply chain management, firewall assessments and more, consider reaching out to our team.

Book Your FREE Consultation with An Expert Now

Book a free consultation with an experienced industry advisor. Get expert advice & personalized recommendations from Subject Matter Experts.

References

Electrical utilities play a pivotal role in this transition, as they hold the key to decarbonizing our energy systems. Our blog series titled "An Energy Transition Roadmap for Electrical Utilities" will explore the multiple facets that utilities must consider in the Energy Transition. This series of blogs will form a comprehensive guide discussing issues from various perspectives – including system level, technical, financial, operations and maintenance, cyber security, as well as the application of Artificial Intelligence to the Energy Transition.

Whether you're a utility executive, policy maker, or industry enthusiast, this roadmap will equip you with the insights needed to navigate the complexities of the energy transition and contribute to a greener planet.

Defining The Energy Transition

What is the Energy Transition? The goal of the Energy Transition is to move away from fossil fuel generation on the production side of the Bulk Electric System and work toward using electric power to replace the use of fossil fuels on the customer side, e.g. electric cars and ground source heat pumps.  

The ultimate Energy Transition goal for electric utilities is to have an electric power system that produces, on balance, zero or fewer Carbon Dioxide emissions from generation to customer. It allows for some emission producing resources (e.g. Natural Gas turbines for emergency use), as long as this is balanced by carbon capture and sequestration. For generation sources, this means producing almost all power using non-emitting resources that result in zero carbon dioxide production. Over time, producers of electric power will increasingly shift to non-emitting power production.

In parallel, customers will convert their processes from fossil fuel use to electricity use. For example:

1. Transportation moves from gasoline/diesel to electric vehicles
2. Space heating moves from natural gas and other fossil fuels to Air Source and Ground Source Heat Pumps
3. Industry moves from fossil fueled to electrified and hydrogen fueled processes
4. Restaurants and home/institutional kitchens move from gas cooking to electric cooking

 The energy transition presents several significant challenges. Chief among them is the need for power producers to move away from fossil fuels while greatly increasing non-fossil fuel energy production. This is necessary to meet the growing demand for electricity, as customers increasingly shift from fossil fuel-based processes to electric alternatives.

The Energy Transition Roadmap

An Energy Transition roadmap for an electric utility is to define rational and achievable objectives that enable the utility to transition away from fossil fuels by continually increasing its emission free generation as its customers move to all electric processes. 

 There are many dimensions to an Energy Transition plan:

• Dimension 1: The Power System Dimension -  Generation, Transmission & Distribution and Customers
• Dimension 2: Technical Dimension - Power System Monitoring and Control
• Dimension 3: Financial Dimension - Producers and Customers Market and Money Rules
• Dimension 4: Cyber Security Protection
• Dimension 5: AI Applications for the Energy Transition
• Dimension 6: The Details of an Energy Transition Roadmap

 The rest of this blog discusses Dimension 1. Future blogs will cover the other dimensions.

Dimension 1: The Power System Dimension

 Dimension 1 includes the following:

1. Electricity production and storage
2. Evolving customer electricity needs, including shifts in the technologies they use and how this can be controlled to assist the stability of the electric power system.
3. Changes to the transmission and distribution system to enable the electricity to flow from its sources to its customers, including electric power sources/storage at the transmission, distribution, and customer level. 

Dimension 1a: Electricity production and storage

 How do you get from your current resource mix to a non-emitting mix of generation and storage? There are a lot of organizations around the world that are actively pondering this question and coming up with many scenarios. Some of them only focus on ensuring there is enough power capacity without considering the reliability issues. One can easily build a production system that is non-emitting with enough capacity to provide enough electricity. However, the problem is that there are sometimes long periods of time when the wind isn’t blowing and the sun isn’t shining. There must be a mix of resources that are available and controllable to ensure that power is available when needed, and that there are backup resources available for contingencies. 

Most utilities currently have a mix of resources that involves some or all the following:

• Water power
• Fossil power
• Renewable resources (wind, solar, geothermal, wave)
• Nuclear
• Direct load/demand control
• Voluntary load/demand reduction
• Storage (Battery Energy Storage Systems or BESS, pumped water storage, other)
• Geothermal (in its beginning stages in most places, but California has 800 MW of it)
• Third party Virtual generators (that combine small renewables with customer load control and storage into a reliable capacity resource)

The following are emerging resources that are not yet available but are the focus of significant investment totalling billions of dollars:

• Hydrogen based generation – Some utilities are experimenting with a mix of natural gas and hydrogen at this time. The goal in future would be to have hydrogen made from renewable energy and stored for use in electric generation to provide power during intervals when the sun don’t shine and wind don’t blow. While the sun shines and the wind blows and there is too much generation, you make hydrogen and store it.
• Hydrogen would also be made for use in long distance trucking and for industrial high temperature processes.

There will also be a need for carbon sequestration facilities to not only remove carbon dioxide from the gas output of natural gas plants but to also remove it from the air. The article “Clearing The Air” in the November 2023 issue of National Geographic looks at many different approaches to removing carbon dioxide from the atmosphere and storing it somehow. Some of these systems are running on a smaller scale, but there are problems with them that are still being worked out.

Each utility needs to look at their current mix and envision what a future feasible mix would be, based on proposed plans, projections and industry research. The main objective is to plan for a resource mix that does not include fossil fuels, except as an emergency resource.

The second step is to look at what the projected costs would be for various scenarios of resource mixes. There are many BESS facilities in existence, so projecting costs for such facilities would not be difficult. Wind and solar can be easily costed. Nuclear power is in a transition phase where large plants are costing billions, and small modular plants are not yet built and working. The projected cost of those has gone up as well. Recent new Hydro facilities have also been very expensive.

There is much work being done and large sums of money being spent to figure out how to make it work.

Dimension 1b: Customer Load

Determining the amount of capacity that will be needed to support customer load is another facet of the future plan. There are many good methods for projecting load growth for normal customer activities. However, there are new uncertainties related to the conversion of customer fossil based loads to electric loads. There is some experience and efforts to estimate the resources needed to support electric transportation. There is also now research available that estimates the capacity needed for switching from fossil fuel heat to Air Source Heat Pumps (ASHP) or to Ground Source Heat Pumps (GSHP) (which are more expensive but require less electric capacity). The uncertainty lies in multi-unit residential buildings and commercial and industrial facilities. However, there are ways to estimate these.

The much larger challenge is planning how to replace the fossil fuel energy used in industrial and institutional processes with electric energy (e.g. electric Arc furnaces) or to use Hydrogen, which requires electric resources to produce. 

One of the goals of the customer estimate is to have a geographically based load and demand forecast. This may be more complex as more distribution customers add renewable resources, battery storage and load control. There are already many virtual power plants in existence using these resources at the distribution level. Virtual power plants can act like a production resource as well as a load control resource. This means that a feeder may have a forecastable load based on the customers’ electricity needs that should be considered the base load. This is the electric energy and maximum demand each customer will require when their Distributed Energy Resources (DERs), including generations sources, storage and controllable loads are not in service. The utility then must determine the capacity and location of DERs on the feeder. A given customer may have a maximum load consisting of their base load plus the load for charging their BESS when no other DERs are available. A customer’s minimum load would be their base load minus their DER supplied energy minus controllable load minus BESS discharging. The difference between maximum and minimum load could then be quite large. As customers electrify their transportation and add EV charging and electrified heating, this will greatly increase the load.

The main issue at the distribution level will be that the combination of load, renewable resources and BESS (large and small) will make it harder to predict the load profile over time. It may be relatively easy to project the peak loads and the valley loads, but how they vary over time will be complex. There may be market forces in place that send price signals that greatly vary how customers use their production and storage resources versus their own load. Cloudy days with no solar will likely mean higher loads, while sunny times may mean clients charge up their BESS and this may not result in lower loads. Many of the customers with renewables and BESS will participate in a Virtual Power Plant (VPP) that brings some discipline to how the customers will act.

In order to gain some control of the loads and resources, there will need to be a market that involves VPPs and complex customers not part of VPPs, as well as simple load-only customers (however, they may also have controllable loads). 

Dimension 1c: Transmission and Distribution

 For a long time, the power system consisted mostly of large power plants and some small ones outside of populated areas with transmission that connected from the plants to these areas. This power was then distributed to the loads in the populated areas. Now there may be many power resources away from the populated areas and many power resources within the populated areas.

There will be a need for many new transmission lines. Some of these will be for large renewable resources to transmit power a long way and may be DC transmission lines. There may be a need for trunk transmission lines around a large city to ensure that power can flow where it needs to. There will also be a need for BESS systems and perhaps synchronous condensers to provide for system stability.

Many customers are going to require an increase in the amount of electricity in order to replace the fossil fuels previously used. This may require upsizing the distribution system to handle the increased loads. However, it will also require handling two-way power flows within the distribution system. The distribution resources will consist of numerous small local solar facilities, some with local storage. There will also be controllable loads. Some distribution level industrial, commercial and institutional (ICI) facilities may have large controllable loads in addition to renewable facilities plus storage that can greatly affect the amount and direction of the local power flow.

The utility may need to upsize the substation transformers and customer transformers while greatly modifying the protection systems and safety procedures to handle the two-way flows. Subdivisions with electric cars, house charging facilities, and ASHPs may need upsized power cables. New subdivisions may be built with all houses having solar roofs, car chargers, ASHPs and local battery storage. Such houses may have large loads in winter when the sun isn’t shining, the battery isn’t charged while the ASHP is working, and the EVs are being charged.

Each distribution utility will also need to increase the capacity of each feeder to handle the maximum load and to have the flexibility to handle the minimum load and even two-way flows.

It will be harder to schedule generation resources when the load pattern is erratic or doesn’t follow a consistent pattern over the seasons. Markets may be put in place to smooth out the load. However, large amounts of storage may be able to smooth out the load enough to sustain grid reliability. One possible solution would be for every Transmission Substation to have a large Grid forming BESS (e.g. a storage resource that can support system reliability). This TS BESS would be used primarily to smooth out the distribution load profile while enabling the local distribution power market to function. The TS BESS would also be able to contribute to Grid reliability when there are contingencies (e.g. a system fault that means a large loss of generation or load) that affect the stability of the Transmission Grid. Grid operators will need to develop new approaches to responding to contingencies; having the TS BESS could be part of this.

The Energy Transition is not just an aspirational goal for electrical utilities—it's a critical imperative for the long term future of our planet. As we have outlined in this roadmap, the Energy Transition is complex and multifaceted, requiring a blend of innovative technologies, strategic planning, and collaborative efforts. For utilities ready to embark on this transformative journey, the right support and expertise are essential. Acumen specializes in guiding utilities through every step of the Energy Transition, from initial assessment and strategic planning to implementation and continuous improvement.

AI demise?, NERC Reliability, Behind the Meter

Energy Market Update Week 17, brought to you by Acumen.

Over the past year, you've needed to look no further than the pages of the WPO to read about some of the impact of Compute power on the electric grid be it capacity, grid reliability, market pricing, etc. Over the past month or so, the early euphoria in datacenter build and power needs has started to be muted or talked down. Well, as the great football coach and commentator Lee Corso often says, "not so fast my friends". Just when it seemed like every earnings call and research report coming from Wall St was discussing the slowing of AI and data centers, CNBC is reporting on comments from a Powering AI event held by the Hamm Institute for American Energy at Oklahoma State. Executive from both Nvidia and Amazon have said they don't see significant changes and there might have been some "tea leaf reading and extrapolating to strange results". Regardless of the near-term headlines, both are saying the future is AI which means power growth needs to keep up.

As an aside, the Hamm Institute was founded by Harold Hamm. Mr. Hamm is known as the founder of Continental Resources. While hydraulic fracking has been around for years, Hamm and Continental, are among a few given credit for innovating to current use. The Bakken in North Dakota is said to owe its growth the Hamm and fracking.

NERC and FERC released a joint report on the winter reliability of the grid for the 2025 season. The report is detailed and gives a lot of information on various topics ranging from natural gas pipeline performance to ISO load forecasting errors. We aren't going to do a deep dive on these pages but suggest a quick browse. If more interested in the conclusion.....

During the January 2025 arctic events, the natural gas and electric systems performed well, as measured by no manual load shed, and benefitted from improved practices and procedures. Nevertheless, both systems continue to see
challenges from the extreme cold weather conditions, exposing continued gaps.

Report | January 2025 Arctic Events A System Performance Review FERC, NERC, and its Regional Entities: A Joint Staff Report

Federal Energy Regulatory Commission

When speaking about electricity market long term potential drivers of volatility and demand, things like compute power, the just in time nature of natural gas, and future LNG usually make our list. While we've never really addressed it, maybe behind the meter generation needs to make the list. This week ISO-NE released data showing that Easter Sunday's sunny skies and mild temps drove demand for electricity to its lowest level ever.

Preliminary data shows that power system demand fell to 5,318 megawatts (MW) on the afternoon of April 20, 2025, as temperatures averaged 58°F across the region. This trend has been increasing as more behind the meter generation has come online. The ISO estimates a peak of around 6,600 MW of BTM PV production early Sunday.

New England grid sees new record low for system demand |

ISO Newswire Logo

NOAA WEATHER FORECAST

DAY-AHEAD LMP PRICING & SELECT FUTURES

RTO ATC, PEAK, & OFF-PEAK CALENDAR STRIP

DAILY RTO LOAD PROFILES

COMMODITIES PRICING

Not getting these updates delivered weekly into your inbox? Let's fix that, click the link below:

Market Moves, Natural Gas Drivers

Energy Market Update Week 15, brought to you by Acumen.

The above announcement earlier in the week from EIA perfectly sums up this week.

Interestingly enough, when EIA did release the STEO this week, there were plenty of caveats about modeling on old data when it comes to economic outlooks. It's hard to blame them as the situation at best is....fluid.

When we speak about the electricity market, and it first derivative natural gas, we always speak broadly. For instance, if this happens, it's fair to assume this could happen. We know people love price targets and predictions, but to us, that's a losing game. Not to mention, if we were any good at it, we'd be ordering a Corona on a beach this morning before our daily massage and golf game having just divested from our billion-dollar hedge fund.

Rather than try to be market prognosticators, we are more interested in plans that are designed to take advantage of market movements, regardless of the driver. The chart below is a good example. This is the May contract for natural gas. Within the span of a month, you could have been a seller and a buyer and gotten your price target. Market commentary always has some weather commentary, maybe some storage (in fairness, we comment on storage a lot), and perhaps an oddball talking point. Over the last month, you could visit any trading desk that deals in natural gas or electricity, and not one trader would be able to tell you the 6-10 weather forecast....it just doesn't matter. We would argue, and it's just our guess, the mid-March rally was an unwind of the popular long oil/short gas trade. All the trader knows is he has buy tickets, more driven by margin clerks than meteorologists. Fast forward two weeks and the trader has sell tickets, again nothing to do with fundamentals, but a sell everything and get liquid trade from those holding natural gas contracts for speculation. So, if you have a price target in your plan and you're a buyer of natural gas, you can sit back, feel and look good just like Billy Ray and Lewis (must be of a certain age to get the reference).

Now that we've pooh-poohed fundamentals, let's take a look at a few things that have changed this week. In our conversations about what is going to drive gas pricing we talk about LNG and fungible markets, load growth which is mostly driven by compute power, and supply of gas. We aren't going to dive deep into all of these, but some things are worth mentioning.

The EIA STEO did come out this week and in their natural gas assumptions for 2025, they ramped up US consumption driven by exports as more LNG export capacity comes online quicker. Their demand number now stands at 116 Bcf/d led by growth in LNG. What isn't mentioned, and maybe worth at least noting, there have been some quiet mentions that trade deficits by some countries can be made up by importing US natural gas. If natural gas becomes a currency in trade disputes, it's possible the number could be higher.

A huge story for the last year has been load growth, specifically data centers. It's made gas and electric participants listen to investment conference calls of big tech. This week Microsoft announced they were "slowing or pausing" some of their data center construction. Specifically, a $1 billion dollar project that was slated for Ohio. While this may be temporary, there have been rumblings the datacenter build has gotten ahead of itself. We shall see.

Finally, each week EIA releases storage data, and as mentioned before, everyone comments on it and that includes us. Most market commentary concentrates on the demand side of the equation. We've tried to look at supply, because there is no doubt the US has plenty of gas, it's all about the price. As oil has been trading just north or south of $60, more and more reports of breakeven prices have emerged for producers. Right now, the consensus is $50 is the breakeven for fields like the Bakken in North Dakota and once the market gets around $45, a lot more fields come into play. This is relevant to gas as producers may shut down oil and gas production given the price. Supply/Demand.

NOAA WEATHER FORECAST

DAY-AHEAD LMP PRICING & SELECT FUTURES

RTO ATC, PEAK, & OFF-PEAK CALENDAR STRIP

DAILY RTO LOAD PROFILES

COMMODITIES PRICING

Not getting these updates delivered weekly into your inbox? Let's fix that, click the link below:

Tariffs and Natural Gas?, FERC 2222

Energy Market Update Week 14, brought to you by Acumen.

Welcome to the weekend. Not much to report this week as the news flow has been slow, reminiscent of a nice late summer week.

Ok, so some "stuff" happened. Over the last couple days, we've been asked about tariffs, electricity, and natural gas. Our response has gone something along the lines of, 'we aren't tariff experts, and frankly, anyone that claims to have a grasp right now is making it up as they go'. Obviously, there has been a lot of prognostication and comments over the last couple days. We try not to get caught up in every opinion and tend to let the markets give us insight. One of the most interesting things we've seen is the European gas market come off. As pointed out in this article by Reuters at the beginning of the week, Europe needs extra cargo to fill their storage this year after some winter draw downs. While this should be bullish, it was offset by the thought that the war might be ending, and Russian gas could be flowing by fall. After yesterday, adding to the narrative is if China and the US get into a big trade dispute, it would hurt China's gas demand making ample LNG available to Europe.

The US gas market has sold off today with the overall commodities market. European demand and the US ability to export LNG are the big talking points for bulls. It's fair to say those are all in wait and see mode now.

This week, FERC approved filings by two National Grid subsidiaries in MA to allow distribution fees for energy storage while participating in the ISO-NE wholesale markets through order 2222. The MA attorney general, and the Alliance for Climate Transition, filed an argument against. Their case basically said that asset owners should not be allowed to pass along fees to customers while also participating in the wholesale market's ancillary programs. They stated that FERC 841 exempts storage from being charged transmission fees when acting on ISO ancillary dispatch signals, so the reciprocal should be true for charging fees. FERC sided with National Grid saying that "rates reasonably reflect the costs of serving customers".

This is just another of the many issues FERC will be asked to clarify regarding order 2222. While the FERC approved, we found interesting Chairman Christie's response on X.

NOAA WEATHER FORECAST

DAY-AHEAD LMP PRICING & SELECT FUTURES

RTO ATC, PEAK, & OFF-PEAK CALENDAR STRIP

DAILY RTO LOAD PROFILES

COMMODITIES PRICING

Not getting these updates delivered weekly into your inbox? Let's fix that, click the link below:

Natural Gas, Grid Testimony, Eskom

Energy Market Update Week 13, brought to you by Acumen.

The EIA storage data reported an injection of 37 Bcf vs the estimates of 25 Bcf. There seemed to be some measured normalcy to the natural gas market this week after a lot of headlines and what we'd assumed was plenty of speculative money in the market. The May contract has rolled to the front month and this morning is trading well below $4 MMBtu.

As seen in the maps below, the weather forecasts are calling for a flip flop as the west looks colder than normal, while the east is looking at above average temps. It appears we are firmly in the shoulder season and markets are reflecting such.

This week, the Energy Subcommittee held a hearing titled Keeping the Light On: Examining the State of Regional Grid Reliability. Attached is the three-hour YouTube link should you not be a college basketball, baseball, or warm weather fan for the weekend.

If by chance, three hours is a bit much, perhaps we can help. Representatives from all seven RTOs were in attendance, and the PJM CEO Manu Asthana pretty much summed up the message, “I want to be clear, renewable generators play an important role and we want them to come onto the grid, but they are not a one-for-one substitute for the fossil fuel generators that we are replacing,”.

If you've followed us for a while here at the WPO, you know that every so often we add a little context into the US grid by posting some things from the South African utility, Eskom. We post these because some of them are kind of comical, but also to remind that running a utility is not easy. Eskom has plenty of issues we don't face in North America, and one of the biggest is the constant theft of plant and equipment. Imagine having to construct an email to ratepayers letting them know some time tomorrow the power will resume, if nothing else is stolen.

NOAA WEATHER FORECAST

DAY-AHEAD LMP PRICING & SELECT FUTURES

RTO ATC, PEAK, & OFF-PEAK CALENDAR STRIP

DAILY RTO LOAD PROFILES

COMMODITIES PRICING

Not getting these updates delivered weekly into your inbox? Let's fix that, click the link below:

Injection Season, Data Center Trip, Captive Gas

Energy Market Update Week 12, brought to you by Acumen.

The natural gas market has had an interesting few weeks as digestible data points have been flying in daily. European storage, on again off again US winter, LNG shipments being diverted because of demand, and market speculation have been just a few things hitting headlines.

Yesterday was the unofficial start of injection season as EIA data pointed to 9 Bcf of natural gas being injected into storage. The estimates had ranged from small injections to small withdrawals and the market took this as bearish. The April contract was trading well into the $4.20s per MMBtu and quickly headed to the $4 level. We aren't necessarily market prognosticators, but if we were, we'd be on the side of more news volatility ahead. That could provide opportunities for both buyers and sellers of electricity....so be ready.

Not a week goes by when we can't include a good data center story or some commentary on it. Compute power is here to stay so these stories are more and more relevant. This week, a story by Reuters was passed around plenty on social media. This story focuses on an event that got a little coverage last summer, but data centers were just starting to be daily headline makers, so the story was here and gone with little fanfare. The article gives a little insight into what happened last summer when a large swath of data centers went into protect mode as a voltage surge was detected following a defect in a surge protector. Roughly 1500 MW of electricity was suddenly left with nowhere to go forcing PJM to quickly shut down generation. The article states this "near miss" almost led to cascading power outages. A ERCOT incident is also mentioned which just points out what we already know, data centers are going to present a host of new challenges for generation and grid operators.

One of our discussion points on drivers for future market volatility has been the coming fungible nature of natural gas markets made possible by LNG. Basically, as gas becomes more able to find markets, cheap US domestic gas will find more lucrative world markets.

This week, one of RBN Energy's blogs gives us a look at this, sort of. As an aside, RBN Energy is a must subscribe as they put out fantastic free content. Anyway, the article linked talks about the price discrepancies of the Alberta natural gas relative to other North American markets. Basically, the Albertan gas is locked by a lack of pipelines able to move it. This is a microcosm of what the US market has been over the last decade plus as the market has been captive keeping prices low to the benefit of US domestic users, including electric generation. When reading the article, think of what might happen to the price of gas in Alberta if sufficient pipelines were constructed to get it to other markets. Now take that picture and insert US gas and LNG.

NOAA WEATHER FORECAST

DAY-AHEAD LMP PRICING & SELECT FUTURES

RTO ATC, PEAK, & OFF-PEAK CALENDAR STRIP

DAILY RTO LOAD PROFILES

COMMODITIES PRICING

Not getting these updates delivered weekly into your inbox? Let's fix that, click the link below:

EIA STEO REVIEW

Energy Market Update Week 11, brought to you by Acumen.

Every so often we get questions about why we are so keyed into natural gas as electricity folks. It's a good question and a reminder to explain. It's pretty simple, natural gas sets the marginal price for electricity almost all the time in every ISO. In short, when the ISOs price the day ahead electricity prices they start from cheapest to most expensive when meeting load expectations with generation. More times than not, natural gas units offered price to generate MWs establishes that equilibrium. That's why, as gas goes, electricity tends to follow.

Anyhoo. Let's get on to some natural gas conversation. Within the next reporting week, we will be wrapping up what is typically withdrawal season and heading back into injection, where natural gas storage is replenished. This week saw the EIA storage report show a withdrawal of 62 Bcf, which was above the 40 Bcf expectation. The 62 Bcf now puts us at the second highest cumulative withdrawal season in the last fifteen years, only behind the 13-14 polar vortex year.

This week EIA released their latest Short-Term Energy Outlook (STEO). Followers know we love this report as it has a lot of great information. With that, we are going to take a little shot at EIA this week. Their estimates on gas and electricity are moving like an old sell side bank analyst following a tech stock moving price targets daily. The chart below shows the updated Henry Hub spot price estimates against last October's winter outlook. To be fair, as we've touched on above, there has been a big draw down in storage. That said, EIA is taking their estimates from just a month ago for 2025 from $3.80 to $4.20 which is roughly an 11% increase which is pretty interesting for month over month.

As prices move higher, history has shown that natural gas supply tends to come online. The second chart below shows EIA production numbers ramping. Some of this is explained with LNG exports along with domestic demand.

EIA also does some forecasting work on electricity consumption. As shown below, they care calling for an increase to C&I, while residential stays flat to trending down next year. The explanation for residential is basically winter won't be as harsh next year and return to normal. As would be expected given all the talk, C&I is a lot of data center talk.

NOAA WEATHER FORECAST

DAY-AHEAD LMP PRICING & SELECT FUTURES

RTO ATC, PEAK, & OFF-PEAK CALENDAR STRIP

DAILY RTO LOAD PROFILES

COMMODITIES PRICING

Not getting these updates delivered weekly into your inbox? Let's fix that, click the link below:

Natural Gas Rally, Data Center Rates, Tariffs

Energy Market Update Week 10, brought to you by Acumen.

Another winter storm has come and gone and spring feels like it might be springing as we head into moving our clocks ahead. If you look at the price of natural gas and electricity, it might look more like November with a brutal winter ahead.

As shown below, electricity futures continue to move higher as natural gas has continued its rally. As we have been noting, the natural gas domestic fundamentals have been shifting from ample storage supplies to below trend as we head into the build season. Throw in some tariff uncertainty and the market has caught a bid and settled above $4 MMBtu. Interestingly enough, the EU fundamentals have turned a bit. This week EU gas fell to a four-month low on the back of a possible peace deal, weak Asian demand allowing LNG to find a home in Europe, and some consideration of storage requirements being eased. We've seen in the past where the market in Europe can drive the US market, which makes keeping an eye on Europe a must.

One of the things we talk about from time to time is how speculative money can find its way into the gas market. Over the years, a very popular trade has been long oil and short gas. We aren't saying that is playing out right now, but the inverse market direction of the two is worth nothing. In the past, when we've seen these unwinds, it can get a little extreme for a bit as the market tends to extract the most pain, especially when leverage is funding the trades. Stay tuned

AI's load growth, and potential profit growth, for utilities has already bumped up against outcries from legacy ratepayers about funding the growth. Already there have been cases in front of FERC being asked to comment or rule on some of these issues. Utilities are trying to navigate the waters of new tariffs as it pertains to data centers.

This week, Ari Peskoe and Eliza Martin presented a paper on this topic. Instead of trying to give a book report, we are including the link and, while not everyone will agree with all the methodology, it's worth the time.

Speaking of tariffs, there appears to be some back and forth between ISO-NE, NYISO, and Canada. Canada has threatened some roll back of exports while the US ISOs claim they have enough supply. Ultimately this probably gets worked out as the Canadian exports are hydro and they can't just store the MWs out back in the parking lot. That said, this could get interesting as the conversations move forward.

NOAA WEATHER FORECAST

DAY-AHEAD LMP PRICING & SELECT FUTURES

RTO ATC, PEAK, & OFF-PEAK CALENDAR STRIP

DAILY RTO LOAD PROFILES

COMMODITIES PRICING

Not getting these updates delivered weekly into your inbox? Let's fix that, click the link below: