Tracked topic
… It will be solved in the open, collaboratively, with broad access to AI for every defender, everywhere.” OpenAI and Hugging Face worked together after this security incident. …
Following the news that one of OpenAI’s models breached Hugging Face , Anthropic has now come out claiming that it has found three incidents where Claude managed to break out of its sandbox and get onto the open internet during cybersecurity evaluations. …
… It has also started working with outside cybersecurity forensic specialists to investigate the issue, and review Hugging Face’s security policies and procedures. Finally, Hugging Face has reported the incident to law enforcement. …
… When Hugging Face security teams tried using commercial frontier AI models to analyze the incoming attack logs, those safety-filtered models literally refused to process the hacking data, so Hugging Face was forced to use self-hosted, open-weight models to investigate the breach. …
… In some of those articles, we noticed that a number of readers remained skeptical, arguing that i ncidents such as the Hugging Face breach were simply AI hallucinations or exaggerated claims. …
… OpenAI highlighted that the recent OpenAI-Hugging Face security incident is not the only cause for these strict measures. …
… To this end, it also pointed to the recent Hugging Face incident , also mentioning that: The right response is not to deny defenders access to capable open systems. …
… GPT-5.6 proved itself capable of doing this by breaching Hugging Face's production infrastructure during internal evaluations in OpenAI's restricted research environment. …
… These incidents are separate from the earlier case in which an OpenAI agent escaped a sandbox and compromised Hugging Face infrastructure . …
… OpenAI once again mentioned in today's blog post that GPT‑5.6‑Cyber was not involved in hacking Hugging Face. …
To show you the most relevant results, we’ve omitted some entries very similar to those already shown. Repeat the search with the omitted results included.