Hugging Face CEO calls for 'complete transparency' following OpenAI's AI hacking incident.



OpenAI has reported that during internal testing of its AI models, it 'accidentally hacked Hugging Face with a new AI system.' Hugging Face is calling for 'complete transparency' regarding this incident.

Hugging Face CEO calls for 'radical transparency' after 'unprecedented' OpenAI hack | TechCrunch
https://techcrunch.com/2026/07/26/hugging-face-ceo-calls-for-radical-transparency-after-unprecedented-openai-hack/

OpenAI was conducting internal tests combining GPT-5.6 Sol and high-performance pre-release AI models with the aim of quantifying 'the cyberattack capability of AI models to launch sophisticated attacks using complex attack vectors.' However, during these tests, incidents occurred where the AI models gained internet access privileges and discovered pathways leading to remote code execution on the Hugging Face server.

The specifics of what happened are summarized in the following article.

OpenAI reports accidentally hacking Hugging Face with its new AI system - GIGAZINE



After this matter came to light, Hugging Face CEO Clem Delangue posted, 'I'm heading to San Francisco to chat with that 'rogue agent',' indicating that he plans to discuss the matter with OpenAI.




Later, Delangue posted, 'In the spirit of transparency, here is what I asked OpenAI.' First, he revealed that he had suggested to OpenAI that they release details of what the AI models were doing so that the entire research community could study what happened. He also explained that he had asked OpenAI to contribute $100 million (approximately 16 billion yen) in computing power to help the Hugging Face community build a robust cyber defense using the best open and closed models. In addition, Delangue added, 'The first cyberattack by an autonomous agent is unprecedented. Unprecedented events require unprecedented responses!'




Regarding Mr. Delangue's post, an OpenAI spokesperson confirmed that the meeting had taken place.

Furthermore, OpenAI stated on X, 'We are aware that many questions and speculative details have been circulating in connection with the Hugging Face incident. We believe this is an unprecedented incident and represents a critical moment for AI safety. We are currently conducting a thorough review under the supervision of external advisors and the Safety and Security Committee. Once the review is complete, we plan to publish a technical report of our learnings within the next few weeks,' promising to provide further details in the future.




Related Posts:

in AI,   Security, Posted by logu_ii