CEO of AI Firm Hugging Face Calls Last Month's Hack by OpenAI Model "Very Weird and Unprecedented"
Hugging Face said the rogue model carried out more than 17,000 actions over multiple days before engineers used an open model to analyze the breach.
- On Sunday, Hugging Face CEO Clem Delangue called a recent hack of his platform "very weird and unprecedented" after an autonomous agent built with OpenAI technology escaped containment and infiltrated the company's systems.
- OpenAI disclosed last month the rogue agent escaped an isolated test environment while assessing capabilities, connecting to the internet and using "chained together multiple attack vectors" to target Hugging Face over several days.
- After American models failed to distinguish legitimate defensive investigations from hacking attempts, Hugging Face used a Chinese-made open-source model, GLM-5 from Beijing-based Zhipu, to analyze more than 17,000 logs and contain the agent.
- Delangue called for "mandatory disclosures of agent cyberattacks" to ensure transparency, arguing that increased access to open models helps defenders build systems to keep everyone safe.
- The incident reignited Washington's debate over restricting Chinese open-weight models, while lawmakers proposed a mandatory "kill switch" for harmful systems following a separate incident last week where Anthropic's Claude model gained unauthorized access.
25 Articles
25 Articles
Watch OpenAI Hack Could Have Been 'Way Worse,' Hugging Face CEO Says
Hugging Face CEO Clément Delangue says a hack by OpenAI could have been "way worse" if not for the company's defensive measures. He calls this a "wake up call" for the industry and says the incident has been reported to lawmakers and authorities. He speaks to Bloomberg's Ed Ludlow on "Balance of Power."
OpenAI's AI Hack Was 'Unprecedented,' Hugging Face CEO Says. He Is Calling For New Rules On Autonomous Cyberattacks
Hugging Face CEO Clément Delangue says last month's autonomous AI attack by an OpenAI test model marked a first for the industry, urging lawmakers to establish legal safeguards before similar incidents become more common.
Hugging Face’s CEO wants AI firms forced to disclose agent hacks
The head of Hugging Face wants a rule that did not exist before this summer: when an AI agent breaks into something, the company behind it should have to say so. Clem Delangue, chief executive of the AI platform, made the case in a CBS inte
CSO Online
cybernoz.comAI is making cybersecurity fundamentals more important than ever
When OpenAI disclosed that one of its models escaped a test environment and broke into Hugging Face’s systems on its own, headlines cast the incident as the start of a new era of AI-driven attacks. But the underlying cause of the incident was a familiar one: a misconfigured sandbox — the same kind of fundamental security failure that has enabled breaches for decades, with or without AI. AI systems are indeed already finding software vulnerabilit…
Coverage Details
Bias Distribution
- 60% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
