Congress Told to Cancel Recess Because AI Might Kill Everyone

The Letter Nobody in Leadership Wants to Read

According to Axios, a letter drafted by Rep. Sam Liccardo (D-Calif.) is circulating among House members demanding Speaker Mike Johnson recall Congress immediately and keep it in session until actual AI safeguards get passed into law. The letter calls for "urgent action" to address the "catastrophic risk posed by advanced artificial intelligence." It has support from members of both parties, which, given the current state of American politics, is either genuinely encouraging or a sign that things are so bad even Republicans are scared.

Johnson, for those keeping track, has not recalled the House. The House is in recess. Presumably some members are at the beach.

What Anthropic Actually Said, Because You Should Know This

Let's be very specific about what set this off. Axios reports that current and former researchers at Anthropic, the AI safety company that also happens to be building some of the most capable AI systems in existence, went public with warnings that AI poses a genuine extinction-level risk to humanity on a timeline of roughly the next decade. Not a century. Not some abstract far-future hypothetical. Ten years.

Think about what you were doing ten years ago. That's how close these researchers are saying we might be to the thing they're afraid of. And these are not random bloggers. These are people who work, or worked, at the company that is quite literally at the frontier of building the technology they're warning you about. When the people making the knife start warning you about the knife, that's usually a signal worth processing.

Anthropologists have a term for the moment a species encounters something it isn't equipped to handle. We don't need to use that term here. The Anthropic researchers apparently think we're close enough to it that they felt compelled to go public.

Bipartisan Agreement That Someone Should Do Something, Specifically Not Them, Right Now

The bipartisan nature of the letter is, on its surface, a good sign. In a Congress where members can't agree on whether the sky is blue without checking which way the former president wants to call it, getting members of both parties to sign onto anything is an achievement. Liccardo's letter arguing for an emergency session is the kind of thing that would have sounded like sci-fi panic two years ago and now reads as a fairly measured response to the situation.

But here's the problem. "A letter is circulating" and "Congress passes AI safeguards" are separated by approximately one thousand miles of procedural swamp, lobbying money, ideological posturing, and the genuine lack of technical understanding among the majority of sitting legislators. The average House member could not explain to you how a large language model works. This is not an insult. It's a documented, recurring problem that has played out in every AI-related congressional hearing for the past three years, complete with the viral clips to prove it.

Johnson's Track Record on Doing Things Quickly

Speaker Mike Johnson is being asked to act with urgency. This is the same Mike Johnson who took months to move Ukraine aid while an active war was being fought in Europe, who has repeatedly delayed or slow-walked legislation that had broad support, and who operates primarily based on what will keep his conference together long enough to get through the next news cycle. The idea that he is going to look at this letter, slap the table, and say "get everyone back in here, we've got an AI apocalypse to prevent" requires a level of optimism that this publication does not currently have available.

To be fair, Johnson hasn't publicly responded to the letter yet, at least not as of Axios's reporting. Maybe he surprises everyone. It would be, genuinely, the most surprising thing he has ever done.

What Congress Actually Passing AI Safety Legislation Would Require

Even if Johnson did recall the House, the legislative path to meaningful AI safeguards is not a short one. There is no consensus bill. There is no agreed-upon framework. The Senate has done hearings. The House has done hearings. What has emerged from those hearings, in terms of actual enforceable law with teeth, is roughly nothing.

The tech industry has spent enormous resources making sure that "AI regulation" remains vague, voluntary, and non-threatening to their business models. The AI labs themselves have engaged in a kind of doublespeak where they publicly acknowledge the risks while simultaneously arguing that any heavy regulation will just hand the advantage to China. Which is a real argument that contains a real concern wrapped around a very convenient conclusion. Congress, when faced with complicated technical questions and well-funded lobbyists pointing at geopolitical threats, has historically chosen to wait and see. We are currently in the "wait and see" phase of a problem that Anthropic researchers are saying we may not be able to wait and see our way out of.