Anthropic Stays Out as Open-Weights Letter Hits 50 Signers

Anthropic is drawing pointed criticism from Silicon Valley peers after declining to sign an open letter defending open-weight AI models, a document that now carries 50 signatories including, notably, OpenAI.

Key points

• An Nvidia-backed open-weights letter doubled its signatories to 50, but Anthropic and Amazon both declined to sign.
• OpenAI’s decision to sign drew particular attention, sharpening the contrast with Anthropic’s absence.
• Criticism is now coming from across Silicon Valley targeting Anthropic specifically as the only major AI lab not supporting open model weights.
• The open-weights debate carries direct policy implications, touching on questions of AI access, safety oversight, and competitive dynamics that regulators are actively weighing.
• Anthropic’s position is consistent with its prior conduct: the company has not released open-weight models and has not publicly explained its absence from this letter.

What the letter is, and who signed it

The letter, associated with Nvidia’s Jensen Huang, advocates for open-weight AI models, meaning AI systems whose underlying parameters are made publicly available for inspection and modification. The signatory count doubled overnight once OpenAI added its name, bringing the total to 50. Meta, which has staked a significant portion of its AI identity on open-weight releases through the Llama series, was an expected supporter of this kind of initiative. OpenAI’s participation was the more surprising development given its historically closed approach to model weights.

Amazon’s absence is notable given its deep financial ties to Anthropic, but the two companies make independent product decisions. The sources do not explain Amazon’s reasoning.

Why Anthropic’s absence is drawing disproportionate fire

Anthropic has never released open-weight models. That is a deliberate choice the company has tied to its safety-first positioning: the argument being that releasing weights broadly makes it harder to control how powerful models are used. That logic has always attracted skeptics who see it as competitive self-interest dressed in safety language, and this week’s letter episode has given those critics a fresh focal point.

Business Insider reports that Anthropic is now being singled out as the only major AI lab not supporting open models. The framing matters. When every other significant lab, including OpenAI, has signed or publicly supported open-weight principles, abstaining becomes a more visible stance than it would have been in a more divided field.

Anthropic has not issued a public statement on its absence from the letter, based on available sources.

Does this affect the IPO story?

For investors tracking Anthropic’s path to a public offering, the open-weights debate is worth watching for two reasons. First, policy. Governments in the U.S. and EU are actively forming rules around AI model access, and where a company stands on open weights will shape how regulators treat it. A lab that opposes open-weight advocacy could face friction in some regulatory environments, or find favor in others that prioritize safety controls.

Second, competitive positioning. Anthropic’s closed-model strategy keeps it dependent on its own distribution channels and enterprise agreements. Open-weight competitors, including Meta’s Llama ecosystem, create alternatives that can undercut pricing or reduce switching costs for enterprise customers. The letter episode highlights that Anthropic is increasingly isolated in defending this approach, at least publicly.

Claude Opus 5, released earlier this week with benchmark results that drew wide attention, demonstrates that Anthropic’s closed approach is still producing competitive frontier models. Whether strong benchmarks are enough to offset the reputational and policy costs of standing apart from the industry on open weights is a question Anthropic has so far chosen not to answer publicly.

This site is independent and not affiliated with Anthropic. Nothing here is investment advice.