Anthropic Tells the Government the Thing Is Coming in Two Years

Anthropic submitted recommendations to the White House's AI Action Plan this week, and buried inside the policy language — the kind of document that is by design unreadable, a gray fog of "stakeholder frameworks" and "responsible deployment considerations" — is a claim so stark it barely registered.

World-changing AGI. No later than 2026 or 2027. Replacing most known human jobs.

Not a think-tank. Not a doomscroller on X. The company. In official government correspondence.

There's a specific genre of cognitive dissonance that only Silicon Valley can produce at this scale — where you simultaneously accelerate toward a thing and warn everyone the thing is coming, and both moves are somehow sincere, and neither slows you down. Anthropic has been doing this longer than most. The company was founded by people who left OpenAI because they were scared of OpenAI, and then built the same product faster.

The OSTP submission is that energy, formalized on letterhead.

What they're actually asking for is government coordination — infrastructure, export controls, workforce preparation, the usual list. All reasonable. All things that would take a decade to meaningfully implement. Which is why the timeline admission is so interesting. They're not saying "we have five to ten years to figure this out." They're saying the train is already visible from the platform and the schedule says two years.

The world replacing most known human jobs in under twenty-four months — not as a sci-fi premise but as the operating assumption inside a policy brief — is a sentence that deserves to just sit there for a moment before anyone starts talking about retraining programs.

Nobody is ready. Anthropic knows nobody is ready. They filed the paperwork anyway.