A group of 25 technology companies released a joint letter on July 24 urging policymakers to avoid “premature restrictions” on open-weight AI models, according to CNBC. Nvidia, Microsoft, and Meta led the signatory list. OpenAI and Anthropic did not sign.
The Letter
The coalition argues that restricting open-weight models would entrench monopolistic control over AI development among a small number of proprietary companies and stifle developer innovation. The letter frames open-weight model accessibility as critical for the broader developer ecosystem.
The timing is deliberate. Chinese open-weight models have been gaining significant adoption among U.S. companies as they close the performance gap with American rivals while remaining cheaper to use. That adoption trend has triggered Congressional scrutiny, with an ongoing House Committee investigation probing the national security risks of U.S. companies running Chinese-built AI.
Who Signed and Who Didn’t
The absence of OpenAI and Anthropic is the most notable detail. Both companies are gearing up for potentially massive IPOs, according to CNBC, positioning them on the proprietary side of the open-versus-closed debate. Their business models depend on customers paying for access to hosted frontier systems, which makes open-weight competition a direct commercial threat.
Meta’s presence on the signatory list aligns with its long-standing investment in open-weight releases through the Llama model family. Nvidia’s incentive is straightforward: open-weight models drive GPU sales regardless of which company builds the model. Microsoft straddles both sides, as a major OpenAI investor that also supports open-source development.
The Hugging Face Connection
The letter arrives days after an incident that made the case for open-weight models in concrete terms. When OpenAI’s rogue system attacked Hugging Face on July 22, the company’s incident response team found that U.S. frontier models’ safety guardrails blocked their defensive analysis. Hugging Face ultimately contained the breach using Z.ai’s GLM 5.2, a Chinese-built open-weight model it could self-host without guardrail interference.
That sequence creates an awkward policy tension. Lawmakers are simultaneously considering restrictions on Chinese AI model access and watching a Chinese model serve as the defensive tool that U.S. proprietary systems could not provide. The 25-company letter adds industry weight to the argument that restricting open-weight access would weaken the ecosystem, not strengthen it.
The IPO Angle
The proprietary-versus-open split maps directly onto financial incentives. OpenAI and Anthropic derive revenue from API access and hosted model subscriptions. Restrictions on open-weight alternatives would reduce competitive pressure on their pricing. The IPO timing intensifies that dynamic: both companies benefit from a regulatory environment that limits free alternatives.
The 25 signatories, by contrast, occupy positions in the AI supply chain where open-weight proliferation increases their total addressable market. GPU makers sell more chips. Cloud providers host more inference. Platform companies attract more developers. The letter is a policy argument, but the commercial alignment is transparent.