New York Gives Utilities 60 Days to Inventory Every AI System in Their Operations

Key Facts
  • New York PSC opened case 26-M-0552 on September 17, 2026
  • Electric, gas and water utilities must file an AI use-case inventory within 60 days, by November 16, 2026
  • Reporting continues twice a year after the initial inventory
  • Utilities must also disclose the policies, procedures and protocols governing each AI system
  • The order names seven risk categories, including hallucinations, misconfiguration errors and cybersecurity attacks

The New York Public Service Commission has ordered every electric, gas and water utility in the state to file a written inventory of each artificial intelligence system running in its operations. The Commission issued the order on September 17, 2026 under case 26-M-0552. Utilities have 60 days to answer, which sets the deadline at November 16, 2026. The docket is titled “Proceeding on Motion of the Commission Regarding an Examination of Artificial Intelligence Utilization in the Operations of Public Utilities and Other Regulated Entities.”

The inventory is only the first filing. Utilities must then report on their AI use twice a year. They must also disclose the policies, procedures and protocols that govern those systems. The Commission said it will measure those protocols against commonly accepted AI governance frameworks. State officials describe the review as one of the first efforts of its kind in the nation.

What the utilities already run

Regulators named working deployments rather than hypotheticals. National Grid has deployed a system called GridCARE to free up interconnection capacity for large-load customers. The New York Power Authority has used AI to read drone-captured data for vegetation management. Consolidated Edison told Utility Dive it applies AI to customer service, early detection of equipment problems, and inspection and mapping work.

The Commission wrote a deliberately wide definition. An AI system is a machine-based system that makes predictions, recommendations or decisions influencing real or virtual environments. The system does this for a given set of human-defined objectives. That wording reaches ordinary machine-learning models for outage prediction and load forecasting, not only generative tools.

Why It Matters

Regulators admit they do not know the size of what they are regulating. The order states that the actual extent and scope of AI systems in New York utility operations remains unclear. A utility that has quietly wired a vendor model into outage restoration or interconnection screening must now put that on the record.

The real exposure sits in the second filing, not the first. An inventory is cheap to produce. Naming the protocols that govern each model invites the Commission to find them thin. Utilities running AI in load forecasting or interconnection study work should expect follow-up questions on model validation. Those outputs already move money in rate cases and queue positions.

Vendors selling AI into New York utilities inherit the disclosure burden. A tool bought as an operations product becomes a regulated input the moment it shapes a reliability or cost decision.

The risk list regulators wrote down

The order names seven failure modes. They are hallucinations, algorithmic biases, transparency issues, data privacy concerns, misconfiguration errors, cybersecurity attacks and functional brittleness. Regulators wrote that these risks carry possible negative consequences for the safety and reliability of New York’s critical infrastructure. Dennis DiBari of the Department of Public Service told the Commission that AI inside critical infrastructure represents a new risk category. Commission Chair Rory Christian called the technology “a double-edged sword for utilities.” The 60-day clock started on September 17.

Critical Perspective

An inventory order is the cheapest thing a commission does. It costs a utility a filing and commits the Commission to nothing. The order sets a 60-day deadline for the first report and a twice-yearly cadence after that, but the record shows no penalty attached to a thin answer. Case 26-M-0552 is an examination, not a rule. Nothing in it yet tells a utility to stop using a model, retrain one, or prove that a vendor system works.

The definition carries the same problem from the other direction. A machine-based system that makes predictions or recommendations for human-defined objectives describes a load forecast written in 1995 as readily as a language model bought this year. A utility reading that wording strictly would file hundreds of entries. A utility reading it loosely would file the three tools with AI in the product name. Both answers satisfy the order as written, and the Commission will have no way to compare one filing against another.

Worth watching: the Commission says the extent of AI in New York utility operations remains unclear, and it has chosen self-reporting to find out. The systems with the most consequence for ratepayers are the ones that shape interconnection screening and load forecasts, because those set queue positions and capital plans. Those are also the systems a utility has the least reason to describe in detail. The November 16 filings are worth reading for what is missing from them.

Sources

Related Coverage

Compliance Impact
✓StatusFiled
⏰TimelineNew York PSC opened case 26-M-0552 on September 17, 2026
⚠AffectsElectric, gas and water utilities must file an AI use-case inventory within 60 days, by November 16, 2026

Related post