Hillary Clinton calls open-source AI dangerous, Delangue pushes back

At the Clinton Global Initiative, Hillary Clinton said open-source AI models are not an "unadulterated good" and pose additional dangers, with Reid Hoffman warning that "one-to-three people in their garage" could find novel attack vectors. Hugging Face co-founder Clem Delangue countered that the real risk comes from secret frontier labs with massive compute, while open models let smaller organizations defend themselves.

• Clinton: open-source AI is not an "unadulterated good" and poses extra dangers
• Hoffman: 1-3 people in a garage with an open model could cause major damage
• Delangue: risk comes from asymmetry of secret labs, not open source
• US Treasury: OpenAI management is responsible for the Hugging Face incident