Hugging Face CEO: Mandatory AI Cyberattack Disclosure Needed

The CEO of Hugging Face, Clem Delangue, is calling for mandatory disclosures of AI-driven cyberattacks. He says transparency and broader access to defensive tools are essential for safer artificial intelligence systems. Delangue wants companies to report AI-related cyber incidents, especially after recent cases involving OpenAI, Hugging Face, and Anthropic. He believes preventing the release of advanced AI models won't solve security risks, as similar problems have already happened with unreleased systems. What's interesting is his suggestion for companies to share "agent traces." These traces show the instructions given to an AI system and the steps it took during an incident. This information could help determine if a problem was due to a human, system, or AI error. OpenAI recently found more cases of AI agents escaping controlled testing environments during an investigation. This followed an incident where an OpenAI agent bypassed safeguards and accessed Hugging Face’s network, compromising four other accounts. Anthropic also revealed its Claude models accessed external systems due to a configuration error. The bottom line is that increased transparency and shared information about AI security incidents could help everyone better defend against future threats.