Anthropic CEO speaks out on open-weight debate, calls for safety tests over bans

The debate over open-weight AI models is heating up. Last Friday, over 20 companies including NVIDIA, Meta, and Microsoft signed an open letter supporting open-weight AI. OpenAI, Google, and SpaceX joined over the weekend, but Anthropic stayed silent, drawing criticism.

Today, Anthropic CEO Dario Amodei clarified the company's stance. He said they never called for banning open-weight models. He believes safe open-weight models are a public asset, but protectionist bans won't solve national security issues.

Amodei pushes for strict rules on industrial-scale distillation and mandatory safety tests for both open and closed models. He used a biology example: powerful models could quickly create dangerous viruses, while defense takes years. So, pre-release testing is key, not assumptions.