Anthropic is finding itself increasingly isolated in Silicon Valley’s AI community as the company remains the only major US frontier lab that has not signed the industry-wide “Open Weights and American AI Leadership” letter, according to multiple reports this week.
The backlash represents a significant shift in the company’s public standing. Once viewed as the thoughtful safety-conscious alternative to OpenAI, Anthropic now faces criticism from former allies—including prominent venture capitalists and startup founders who argue the company’s refusal to support open-weight models amounts to a regulatory capture strategy designed to entrench closed AI systems.
“They positioned themselves as the safe choice, the responsible choice,” said one prominent venture capitalist who requested anonymity. “But when given the opportunity to support American AI leadership through open innovation, they chose to align with the most restrictive regulatory approach instead.”
David Sacks, the venture investor and former PayPal executive, has been particularly vocal in his criticism. At recent industry events, Sacks has argued that Anthropic’s stance reflects a deliberate strategy to leverage regulatory relationships to block open-source competitors—positioning the company’s safety commitments as a competitive weapon rather than a genuine principled position.
The divide comes at a critical moment for the AI industry. While OpenAI, Meta, Google, and Mistral have all released open-weight models of varying capabilities, Anthropic has doubled down on its closed approach. The company has argued that frontier models are too dangerous to release openly and that open weights would accelerate harmful AI development.
But critics note that Anthropic’s own research has also contributed significantly to the open-source ecosystem through publications and the Claude Artifacts program—leading some to question whether the company’s position is principled or simply competitive.
“They publish papers about constitutional AI and RLHF techniques that help make models safer, but they won’t release the models themselves,” noted one AI researcher at a major tech company. “It’s hard to square that with their stated mission of benefiting humanity.”
The tension reflects broader fractures in how the AI industry views the open-closed spectrum. OpenAI, once the champion of open research, has progressively tightened access to its frontier models. Meta has positioned itself as the open-source champion, releasing Llama with increasingly permissive licenses. Anthropic has occupied a middle ground—more open than OpenAI in its research practices, but firmly closed in its model releases.
The timing of the backlash is notable. Anthropic is reportedly preparing for an IPO, and maintaining relationships with the Silicon Valley investment community will be critical to that process. The growing criticism could complicate the company’s path to public markets, especially as investors increasingly ask difficult questions about competitive positioning and regulatory risk.
Anthropic has not publicly responded to the criticism beyond restating its safety-focused mission. The company continues to argue that responsible AI development requires careful control over who can access the most capable models—a position that resonates with some policymakers but increasingly alienates the open-source community that has driven much of the industry’s recent innovation.
As the debate continues, one thing is clear: the AI industry’s “open versus closed” debate is no longer just a technical discussion about model capabilities. It’s become a fault line that defines competitive strategy, regulatory relationships, and fundamental questions about who should control the most powerful AI systems being developed today.