Leading AI labs OpenAI and Anthropic, despite competing for customers, are aligning in Washington to warn policymakers about the risks posed by powerful Chinese open-weight AI models. This shared stance contrasts with researchers and startups who advocate for broad access to AI, arguing it's crucial for competition and scientific progress. Closed-model labs like OpenAI and Anthropic could benefit from increased scrutiny on open-weight models, a move critics, including David Sacks, label as potential regulatory capture that could entrench dominant companies.

Anthropic CEO Dario Amodei highlights the safety challenges of open-weight models, as developers lose control over access and updates once weights are released. Conversely, supporters see this lack of central control as a strength, preventing single-entity monopolization. OpenAI views open-source as a means to "democratize access" within a national framework. The company has also raised concerns about "distillation," where models are trained on existing proprietary models, a practice echoed by U.S. officials like Jamieson Greer and Scott Bessent, who view it as intellectual property theft.

Advocates of open models argue they are vital for enterprise diversification and scientific discovery. Suresh Venkatasubramanian, a former Biden White House tech adviser, expressed concern that restrictions on Chinese models would hinder research. While OpenAI and Anthropic share regulatory concerns, their approaches differ, as seen in their endorsements of a Massachusetts AI safety bill, with Anthropic favoring a more stringent version requiring independent testing and stronger enforcement.