Dario Amodei, CEO of Anthropic, stated on Monday that he has never supported a ban on open-weight AI models, countering claims that his company is attempting to protect its closed-model business. Amodei's comments come as Anthropic stands apart from a collective industry effort, which includes companies like Nvidia, Microsoft, Meta, Google, and OpenAI, advocating against restrictions on open-weight AI technology.
The industry push was prompted by the introduction of Kimi K3, a Chinese open-weight model that has demonstrated performance comparable to U.S. systems at a lower cost. The Trump administration has considered actions against open-source models, including potential sanctions on Chinese laboratories accused of appropriating U.S. AI research.
In a blog post, Amodei emphasized that he views models lacking dangerous capabilities as beneficial to the public. He argued that banning Chinese open models would not effectively mitigate threats, as he believes that malicious actors are unlikely to be legitimate U.S. businesses. While acknowledging that such a ban could protect U.S. companies like Anthropic from competition, he insisted that this was not his intention.
Amodei expressed a desire for more targeted regulations rather than an outright ban, suggesting three specific policies: stricter controls on advanced chips and manufacturing equipment sent to authoritarian regimes, a crackdown on industrial-scale distillation, and mandatory safety testing for capable models, regardless of whether they are open or closed. He noted that Anthropic faced criticism for not joining the open-weight letter initiated by Nvidia's CEO, with Google and OpenAI eventually signing on after initially abstaining. Amodei's absence from the letter has led to accusations that his company is leveraging safety concerns to defend its business model.