The open-weights debate has taken a new turn — and this time Anthropic is standing pretty much alone. As Washington weighs restrictions on some Chinese models, OpenAI, Google, Microsoft and SpaceX signed an open letter urging the U.S. government to support open-weight AI. Anthropic didn’t sign. And the criticism from across Silicon Valley has been piling up.
The only major holdout
Axios sums it up in an analysis titled «Anthropic’s lonely island»: Anthropic is the only major AI lab that hasn’t backed the defense of open weights. That’s no accident — it fits a line the company has held for months.
No lab has pushed harder for regulation than Anthropic. The company has backed state-level AI laws and proposed an FAA-style regulator to screen advanced models before release. CEO Dario Amodei has long warned that developers lose control of frontier models once their weights are out in the open.
Not just about open weights
The striking part of the Axios piece: this isn’t only about one letter. Anthropic is growing more isolated across nearly every major policy fight — from military use to AI guardrails. Where other labs pragmatically close ranks, Anthropic is more willing to be openly confrontational, even when it’s intellectually closer to chief rival OpenAI than it looks from the outside.
My take
You can read this two ways. The critical version: Anthropic is boxing itself into a corner and risks losing its voice in the debates that matter. The charitable version: this stance is Anthropic’s whole brand. A company with «AI safety» in its DNA has to be inconvenient sometimes — otherwise the label means nothing.
I lean toward the second reading, with a question mark. Sticking to your principles is admirable. But policy gets made by the people in the room. If Anthropic is too often the only one saying «no», there’s a risk the music plays on without them. The interesting question is whether the company can hold this position — or whether the pressure from the industry eventually gets too big.
Sources: