Collective Action for AI Safety: NYC, June 4

The format is simple: a coordination game, an open discussion of the collective action problems related to AI safety, and a quick talk by Alex Bores. Bores helped write the RAISE Act, a well-regarded piece of AI regulation, and faces opposition from a major pro-AI PAC that wants to take credit for his defeat if he loses, thereby chilling similar moves. Anyone in or near NYC is welcome, and the invitation is open, so bring people who’d find it interesting.

The order was lobbied against hard, canceled, then signed about a week later with essentially one change: the pre-release window shrank from 90 days to 30. David Sacks opposed it from its first announcement, reportedly lobbied against it, but is now telling reporters it’s fine. Even Dean Ball, who thinks the order is a mistake, described it as a clear win for the safety side and a loss for the Sacks and accelerationist wing.

For anyone organizing around AI, the takeaway is bigger than the order. The “no oversight whatsoever” position has lost, and it lost in the friendliest possible venue it was ever going to get. The premise that the government has some oversight role, commonsensical but resisted as hard as the accelerationist faction could, has now been conceded by their own biggest political allies. So the seal is broken—everything from now on is a negotiation over the degree and type of democratic (or undemocratic) accountability.

I’ll be in San Francisco for the next couple of weeks for LessOnline and Manifest, meeting people, fundraising, and volunteering with the Doom Debates crew. If you’re attending either event or you’re in the Bay Area and want to talk about coordination, organizing, or how the AI policy landscape evolves from here, get in touch: jordan@spartacus.app.