OpenAI and Anthropic Unite Against Open-Weight AI Risks

OpenAI and Anthropic compete fiercely for customers and trade blows over benchmarks. But in Washington, the two leading AI labs are finding common ground — and it's one that could reshape who gets to build and use the most powerful AI systems in the world.

Rivals Find a Common Enemy

Both companies are now warning policymakers about the risks of powerful open-weight AI models, arguing that once model weights are publicly released, developers lose the ability to revoke access, update safety guardrails, or prevent misuse.

The shared position was highlighted in Axios reporting today, and it pits the two leading U.S. AI labs directly against researchers, startups, and open-model advocates who argue broad access to AI is essential for competition and scientific progress.

Anthropic CEO Dario Amodei has been the most vocal. He argues that open-weight models are harder to keep safe precisely because no single entity can control them. For open-model advocates, that's the point: no company should be able to decide who uses AI or how.

The Regulatory Capture Question

Critics — including Trump administration adviser David Sacks — warn that greater scrutiny of open-weight models could amount to regulatory capture. Rules intended to improve AI safety could instead entrench the largest companies by making it harder for competitors to release models.

The tension is playing out against growing concern over Chinese AI labs. Treasury Secretary Scott Bessent said Tuesday that models from China need to be held to the same standards as U.S. models. U.S. Trade Representative Jamieson Greer went further, calling Chinese distillation — training a model on another company's outputs — a form of IP theft.

Open Weights vs. Open Debate

OpenAI, which has its own open-source model, tried to walk a careful line. "Advances in Chinese open-weight models are not an argument against openness," a spokesperson told Axios. "They reinforce the need for a coherent national framework that enables the U.S. to evaluate new models quickly, manage risks, and get the most powerful AI tools into the hands of cyber defenders."

Suresh Venkatasubramanian, former Biden White House tech advisor, offered a different perspective: "As a researcher, I'm very worried if we get access cut off to all the Chinese models. That would be a big problem for doing research."

State-Level Divisions

The alignment in Washington doesn't extend everywhere. OpenAI and Anthropic diverged this week on a Massachusetts frontier safety bill. Anthropic endorsed a version requiring independent model testing every six months with public reports and civil penalties for non-compliance. OpenAI endorsed a weaker version without those provisions — a reminder that even allies have different thresholds for regulation.

What to Watch

OpenAI CEO Sam Altman will meet with lawmakers and the White House next week to discuss upcoming model releases. With the White House frontier AI framework expected before August 1, and DeepSeek V4 and Kimi K3 open-weight releases landing this week, the debate over who controls AI access is about to get much louder.

Key takeaway: The two biggest names in proprietary AI are now aligned in Washington against open-weight models. Whether that's genuine safety concern or strategic regulatory capture depends on who you ask — but either way, it will shape AI policy for years to come.