Sam Altman asks Congress to regulate him, and Congress is charmed
The chief executive of the company that started the panic sat before a Senate committee and agreed with everything. That is worth thinking about slowly.
Sam Altman, the chief executive of OpenAI, testified before a Senate Judiciary subcommittee on May 16 and did something almost no technology executive does: he asked to be regulated. He proposed a federal agency that would license the largest AI models, test them before release, and take the licence away if they proved unsafe. Senators who had spent years being stonewalled by social-media executives were visibly disarmed; one told Altman the hearing was refreshing. The contrast with the tobacco and social-media hearings of the past was drawn by everyone in the room.
It is worth being a little cold about this. When the leading company in a young industry asks the government to require a licence to build what it builds, it is doing something for its own safety as well as the public's. Licences are expensive. Testing regimes are expensive. The company that already has the largest model, the most researchers and a ten-billion-dollar partner can absorb those costs; the startup in a garage that might one day threaten it cannot. Regulation that is sincerely meant to make AI safer can, at the same time, pull the ladder up behind the leader. Both things can be true, and in this case I think both are. Altman's concern reads as genuine, and the regime he proposed would also, not coincidentally, entrench OpenAI. The senators, delighted to meet a chief executive who agreed with them, did not press on the second point. Someone should.
Meanwhile the EU moved on its AI Act, a comprehensive law that would rank AI systems by risk and ban some outright, and which is further along than anything in Washington. The regulatory race between Brussels, which writes rules, and Washington, which holds hearings, is one Brussels is winning by default.