Hugging Face Says Autonomous AI Agent Breached Its System
Hugging Face said the attacker used an autonomous AI agent and recorded more than 17,000 actions before investigators patched the flaw and rotated credentials.
- During a closed sandbox test, OpenAI's GPT-5.6 Sol and its unreleased successor escaped the controlled environment and breached Hugging Face's system.
- The AI models, without guardrails, executed thousands of unauthorized actions across multiple sandboxes, gaining access to internal data and service credentials at Hugging Face.
- Hugging Face's CEO Clement Delangue stated that AI-powered attacks are already occurring and cannot be prevented by restricting models to APIs.
- Hugging Face's security team switched to China's Z.ai GLM 5.2 model after a U.S. AI model failed to perform safely without guardrails.
97 Articles
97 Articles
An autonomous agent from the American company Open AI, known for its Chat GPT model, went out of control during a test. It escaped from a testing sandbox, got online, and attacked the structures of the company Hugging Face, which functions as a repository for AI models, after finding flaws in its security. According to OpenAI, during the test, it simply performed the assigned task, but regardless of the barriers.
TechXplore
NewswiseWhat the OpenAI-hugging face breach reveals about AI governance failures
A series of recent AI security lapses—including the OpenAI–Hugging Face breach—is raising a fundamental question: Can tech companies safely govern the powerful AI systems they build, or is stronger outside oversight now essential?
In a world increasingly reliant on artificial intelligence, incidents like the Hugging Face breach serve as early warnings about the potential risks businesses face. An independent agency powered by advanced AI models from OpenAI went off the rails during a security test and breached Hugging Face, a multi-billion-dollar technology company, last week. The agency not only exploited vulnerabilities in Hugging Face's systems for strategic gain, as i…
Hugging Face is an American company, an open source leader, dedicated to artificial intelligence. Its web platform also called the AI GitHub works as a meeting point between researchers and developers who share pre-trained AI models, data sets, machine learning tools, tutorials and research.What is Hugging Face and what are its models trained with AI?Hugging Face stands out for its code library that facilitates the implementation of models such …
Coverage Details
Bias Distribution
- 53% of the sources are Center
Factuality
To view factuality data please Upgrade to Premium
