Hugging Face CEO on AI Safety: Accelerating Forward! Successfully Conducted Defense Attempt Using Open-Source Models, Larger Vision Beyond Risk Must Not Be Overlooked

According to Dynamic Beating monitoring, Hugging Face CEO Clem stated that now is not the time to slow down, but to accelerate! Recent AI-driven cyber attacks have sparked discussions about AI's risks. While it is important to have these discussions, we must not overlook the bigger picture. If we work towards it, AI will make the world a safer place, much like most major technologies.

We have seen a glimmer of hope: AI (specifically an open-source model) has successfully defended itself. The same set of systems that helped thwart AI-driven cyber attacks can now assist in defending against millions of attacks every day, while helping us identify and patch vulnerabilities before attackers can exploit them. To achieve this, I believe we need three things:

1. Increase transparency, mandate tracking and event disclosure for proxy network attacks;

2. Continue to criminalize AI-driven cyber attacks and impose meaningful penalties as a deterrent;

3. Equip defenders with the best AI, especially open-source models, to reduce the asymmetric capabilities between attackers and defenders.

If these three things are done well, AI will not only pose new cybersecurity challenges but will fundamentally and meaningfully strengthen cybersecurity.

In the recent OpenAI model breach at Hugging Face incident, bipartisan legislation "AI Emergency Shutdown Act" was introduced by U.S. lawmakers, requiring developers of advanced/cutting-edge AI systems to maintain technical capabilities to throttle, pause, or shut down systems; and authorizing the Secretary of Homeland Security (DHS), in consultation with the Secretary of Commerce and the Director of National Intelligence, to intervene in case of model runaway or catastrophic risk.