Mike Johnson: AI safety is the companies' job, not Congress's
The labs asked Washington for permission to slow down on Saturday. Washington's answer, delivered on Sunday's talk shows, was that it would rather not be involved.
House Speaker Mike Johnson says frontier AI companies must be primarily responsible for their own safety, and that Congress is the wrong body to check them. Appearing on NBC's "Meet the Press," Johnson called AI regulation "a very important issue" but insisted "it's not one that we can rush into," citing the risk of losing the China race — the same argument President Trump made hours later in Ireland, where he said "whoever wins AI wins" and dismissed the warnings as coming from "negative forces" raising "things that won't happen." Johnson's sharper claim was about competence rather than timing: that Congress is "obviously less qualified than the people who are pushing this frontier to know the ins and outs of it," and that the CEOs calling for pacing "don't themselves even have the answer as of today." His plan is a meeting, not a statute — he said he would "summon the platform providers all to one big meeting and work this out together in the same room." That is the industry's own ask, inverted. Amodei and Altman want a binding multilateral framework with independent evaluators inside the labs; Johnson is offering a photo op plus self-policing, and the practical effect is that the pacing decision stays with the two firms that published the essay. Watch the gap between the rhetoric and the calendar: House lawmakers are circulating a letter to bring the chamber back early for AI safeguards, and Rep. Anna Paulina Luna is calling for a special session — pressure enough that "no rush" becomes a position that has to be defended by name. We tracked the industry half of this shift two days ago — Altman matches Amodei's evaluator pledge, Musk says 'Dario is right' — and Obama's intervention on the Democratic side this morning — Obama tells Jeffries: make AI the Democrats' first fight.
Jacob Coxon, the ex-Anthropic researcher whose resignation letter started this week's panic, now says Congress should let the labs self-regulate until a real regulator exists. Coxon, also on "Meet the Press," said the lab executives' recent statements are "completely genuine" and that he would "allow that the current labs regulate themselves" while lawmakers build a long-term framework, arguing any statute has to be able to react to new models as they ship. He also lowered the temperature on the favourite legislative fix: a national kill switch "probably would work on a lot of AIs for now," he said, but there are "a lot of switches," and it stops working in the scenario Amodei described this week — an agent swarm already loose on the internet. It is the most honest version of the insider-warning position, and the least convenient: the person who made the accusation is now endorsing the accused as interim referees.
Hugging Face volunteered to be the auditor Amodei asked for, and the cybersecurity trade is pushing back on who gets the job. Co-founder Clément Delangue announced the Open Alignment Initiative, led by Thomas Wolf, and asked to be included in the "embedded evaluators" program Amodei committed to — turning the request for outside oversight into a business line for the model hub. The friction is over METR, the small nonprofit both Anthropic and OpenAI have used for incident reviews and which Amodei floated as a template evaluator. Twenty-five-year industry veteran Greg Linares called the METR-authority narrative "dangerous, deceptive, and morally corrupt," arguing labs should hand this work to experienced incident-response firms and publish the prompts and evidence; Florian Roth's objection to the July sandbox breach was simpler — "what was actually enforcing the isolation?" Meanwhile Anthropic's frontier red-team lead Logan Graham conceded the substantive point on the record, saying unmonitored AI "does just pose a national security risk." An oversight regime is only as strong as the referee's résumé, and nobody has agreed on one yet.
What to watch: whether Johnson's "one big meeting" gets scheduled before the midterms, and whether any lab names an evaluator other than METR.
If the companies writing the safety rules are the ones being graded, is that self-regulation or self-certification? Tell us in the comments.
