Anthropic's Concept of a Plan, No Actual Plan
The safety-conscious AI company releases a paper on 25% unemployment with vague proposals—more performative hand-wringing from an industry that talks big and does little.
Anthropic has a concept of a plan for 25% unemployment. Notice the word "concept." It is doing a lot of work in that sentence, because what the company does not have is an actual plan. Last week, the same firm made headlines for deliberately hobbling its Mythos model to make it "safe" for research—a move I called safety theatre at its finest. Now it publishes a paper titled with all the gravitas of a student who missed the deadline, acknowledging that AI might put a quarter of us out of work, but offering only a shrug and a promise to think about it some more.
The paper, helpfully summarised by Gizmodo, states that "some displacement may be an intrinsic consequence of AI" and that Anthropic must "prepare and respond." Prepare how? Respond with what? The document outlines proposals for different levels of unemployment but commits to precisely nothing beyond what was already announced. It is the corporate equivalent of saying "thoughts and prayers" while the house burns down. And let us be clear: this is the same company that has raised billions of dollars, the same company that claims to be the responsible alternative to OpenAI, the same company that publishes its constitution like a holy text. Yet when faced with the most obvious societal risk, it produces a paper that even Trump would recognise as a "concept of a plan."
Meanwhile, the real world does not wait for concepts. The workers who are already being replaced are not consulting Anthropic's paper for comfort. They are not waiting for a vague set of proposals. They are updating their CVs. And the firms doing the cutting are not citing the need for a careful, measured approach—they are wielding AI as a convenient excuse for cost reduction. Anthropic's paper exists in a vacuum, as if unemployment is a hypothetical future event rather than a present reality for tens of thousands.
It gets worse. The paper is released alongside news that Anthropic is partnering with Tata Consultancy Services to scale enterprise deployments. So on one hand, the company is actively helping large corporations deploy AI that will displace workers; on the other, it produces a paper fretting about displacement. This is not hypocrisy—it is business as usual. Anthropic can afford to wring its hands because the pain is not being felt in its own offices. The company employs a few hundred people, all of whom are highly skilled and unlikely to be automated any time soon. The 25% unemployment scenario is an abstraction, a thought experiment to be workshopped at conferences.
The phrase "concept of a plan" was famously used by Donald Trump to describe his health care proposal, and it was rightly mocked for its emptiness. Anthropic has now borrowed it for AI policy. That is about as substantive as the rest of the document. There is no timeline, no budget, no mechanism. Just a lot of language about "responsibility" and "preparedness" that sounds profound until you try to pin down a single commitment. For a company that has built its brand on safety, this paper is a remarkably low-effort attempt to appear engaged.
Last month, I wrote about the Altman bomb plot—the real-world violence that can result from the hype these companies create. Now we have a paper that treats mass unemployment as an academic exercise while the CEOs who cause it parade around Davos. Anthropic's concept of a plan is exactly what you would expect from a company that cripples its own models for show: performative, self-congratulatory, and utterly useless to the people who actually need a plan. The rest of us will have to make our own.