Amodei told lawmakers that open-source AI is moving down a "dangerous path"

- Dario Amodei is the CEO of Anthropic, the AI company behind Claude (Anthropic's large language model). He previously worked at OpenAI and was a key figure in its safety-focused split. - This tweet refers to Amodei testifying before U.S. lawmakers (likely a Senate or House hearing on AI policy) where he argued that releasing powerful AI models as open-source code — meaning freely available for anyone to download, modify, and run — poses serious risks. - The "dangerous path" claim reflects a major ongoing debate: companies like Meta (with Llama) and Mistral release open-weight models, arguing for transparency and democratization. Anthropic and OpenAI generally favor controlled, API-based access, warning that open-source AI could be used for bioweapons, disinformation, or other harms without any safety guardrails. - This is part of a broader policy fight over AI regulation, where lawmakers are weighing innovation against national security and public safety concerns.