New York Mandates AI Transparency for Utility Providers

The New York Public Service Commission has issued a direct order to major utility providers, including National Grid, to disclose all instances of artificial intelligence usage within their operations. This regulatory move, announced in Watertown, requires companies to report their AI integration strategies within 60 days. The commission argues that as these technologies become embedded in critical infrastructure, the public and regulators must have a clear view of how they function.

Utilities are increasingly deploying AI for core functions such as customer service interactions, predictive maintenance of power lines, and safety inspections of facilities. While these tools promise efficiency, the commission highlights significant downsides. These include the potential for algorithmic bias, data privacy violations, and what experts describe as functional brittleness, where systems fail unexpectedly under unusual conditions. The order seeks to mitigate these risks by forcing transparency before further adoption.

Regulators target specific AI risks

The inquiry specifically targets known weaknesses in large language models and automated systems, such as hallucinations, where the software generates false information with confidence. The commission also flagged cybersecurity vulnerabilities and misconfiguration errors as primary concerns. By demanding a detailed inventory of AI applications, the PSC aims to identify where these technologies might compromise the reliability of the power grid or the security of customer data.

This approach reflects a broader shift in utility regulation, moving from simple performance metrics to oversight of the underlying decision-making processes. The trade-off for companies is increased administrative burden and potential legal exposure, but the commission insists that the safety of the electrical grid outweighs the cost of compliance. Utilities must now demonstrate that their AI systems are robust, unbiased, and secure enough to handle critical infrastructure tasks.

Mandatory policy disclosure requirements

Beyond listing specific AI tools, utilities are required to submit the policies, procedures, and protocols that govern their use. This includes how human oversight is maintained and how errors are detected and corrected. The commission will review these submissions to determine if additional protective measures are necessary. According to reporting by GN technics/ai (en-US), this step is crucial for establishing a baseline of accountability in an industry where failures can have widespread consequences.

The 60-day deadline creates immediate pressure on utility firms to audit their internal systems. Many may find that their current documentation is insufficient for regulatory scrutiny. The commission’s evaluation phase will likely lead to new standards or restrictions on how AI can be deployed in the energy sector. This regulatory action sets a precedent for other states that may look to New York’s framework as a model for managing the risks associated with automated decision-making in critical services.