New York now requires utilities to report their AI use

What happens if the AI running your power company's outage predictions gets it wrong? New York regulators just decided that's a question worth tracking closely, and they're making utilities show their homework twice a year to do it.

The New York State Public Service Commission took up the vote on Thursday that would compel state utilities to submit semi-annual reports on their AI operations. The move comes as one of the first of its kind nationally, considering how silently the AI technology has started operating at the utilities.

AI is increasingly being adopted by the utilities to perform inspections, manage outages and even in customer service activities. "It's no secret that some of the work already has resulted in benefits to New Yorkers, but it’s equally true that any new tools come with risks," PSC chairman Rory Christian said.

Among the issues of concern are problems relating to privacy and the potential for fabrication by AI models. Dennis DiBari from the Department of Public Service described the increasing use of AI technology in critical infrastructure as being an entirely new threat that needs to be considered by the commission, something that did not even exist a few years ago.

Utilities are required to explain how they are using AI technology and also the kind of AI model being used; this could be either classical machine learning or generative AI models. The first filings are due in 60 days.

National Grid, one of the state's major utilities, said through a spokesperson that it's reviewing the inquiry and will provide the requested inventory, while pointing to its existing policies for safeguarding customer data and meeting regulatory requirements.