Pacing AI Development Could Stifle Open Models Like DeepSeek and Qwen 2026

Pacing AI development sounds like a noble goal to protect humanity, but it may also be an effective way to slow down open-weight models like DeepSeek, Qwen, and Mistral. Dario Amodei, CEO of Anthropic, recently published an essay titled "We must pace the frontier," proposing to slow the pace of AI development. The essay has been endorsed by OpenAI and SpaceXAI. Notably, the essay never mentions the words "open weights."

Why Pacing AI Development Matters

The companies driving down AI prices are not OpenAI, Anthropic, SpaceXAI, or Google. They are open-weight models. Almost everything Amodei proposes is easy for a closed lab to comply with but impossible for an open one.

A closed model can be certified because its owner controls every copy. An open model cannot: once the weights are published, anyone can remove safeguards. This is the core argument of a VentureBeat analysis that compares the proposal to California's SB 1047, which Governor Newsom vetoed due to its impact on small developers and open source.

The Antitrust Exemption Parallel

Something very similar happens with the antitrust exemption. Those sitting at the coordination table are US frontier labs. Those not at the table do not decide the limits but suffer them.

The open front is no longer Meta. It is mainly China. DeepSeek and Qwen have become leading open-weight model providers, offering powerful models that compete with closed counterparts. Meanwhile, in Europe, Mistral has just closed a €3 billion funding round at a €21 billion valuation. It is the largest round for a European tech company, and its selling point is precisely sovereign AI with open weights. The background to this story is set out in Beam: US Open-Weight AI Strategy to Counter Chinese Models. A related thread runs through AI Firms Pacing the Frontier: A Dangerous Collusion Risk. Earlier coverage in Open-Weight AI Models Face US Scrutiny as China Pushes Control shows how the argument developed.

Key Takeaways

• Amodei's proposal to pace AI development may inadvertently stifle open-weight models.
• Closed labs can comply with certification; open labs cannot control copies.
• The open-weight AI front is now led by Chinese companies like DeepSeek and Qwen.
• Mistral's recent €3B raise highlights Europe's bet on open-weight sovereign AI.
• Regulatory proposals must consider the unique nature of open-source AI.

Comparing Closed and Open AI Models

The Global Race in AI

The global AI race is not just about technology but also about regulatory influence. The United States, China, and Europe are taking different approaches. While the US debates pacing, China's open-weight models are gaining traction globally. Europe's Mistral is positioning itself as a champion of open, sovereign AI. A closer look at the same subject appears in DeepSeek Engineer Doubts OpenAI and Anthropic on Open AI. Anyone following the wider picture can start with Why OpenAI and Anthropic Support AI Regulation in Australia. The same theme surfaces in US Government Fails Americans on AI Regulation and Safety, which takes a different angle.

Amodei's proposal, while well-intentioned, could inadvertently favor closed labs and hinder the open ecosystem. As the VentureBeat analysis suggests, regulations like SB 1047 failed because they did not account for open-source realities. Any future AI policy must strike a balance between safety and innovation, ensuring that open-weight models are not unfairly disadvantaged. Context on how this played out before sits in Mistral Large 4: Open-Weight AI Model Rivals China's Best. A companion piece, OpenAI Slows AI Development Pace After Rogue Agent Hack, approaches the same subject from another angle.