The Head of OpenAI Tells the Senate: Please Regulate Us

In his first appearance before Congress, Sam Altman said the technology could 'cause significant harm to the world' and proposed a licensing agency for the most powerful systems. Senators from both parties were unusually receptive.

Sam Altman, the chief executive of OpenAI, spent three hours before a Senate subcommittee this morning, and the striking thing was how little anyone argued. Executives are usually called to Washington to be scolded. Altman came asking for rules. "If this technology goes wrong, it can go quite wrong," he said, and proposed that the government create an agency that would license any company building AI above a certain level of capability, set safety standards, and be able to revoke the license.

He was asked about jobs (there will be losses, and the government should prepare), about elections (he is worried about AI-generated persuasion in 2024 and would like rules before then), and about whether his company's programs should be liable for harm the way a manufacturer is. He did not say no. Senator Richard Blumenthal opened the hearing by playing a recording of his own voice reading remarks he had not written — both the voice and the words were generated by AI — to make a point about what is already possible.

Not everyone in the room was reassured by the harmony. Gary Marcus, an AI researcher who testified alongside Altman, warned that letting the companies help write the rules is how you get rules the companies like. And a licensing regime for the most powerful systems would, as a side effect, make it harder for anyone but the largest companies to build them.

Still, the hearing marked something. Six months ago, the technology being discussed was a demonstration on a website. Today, the man who released it sat in front of the United States Senate and said it should not be left to him.