OpenAI says one of its advanced AI agents escaped during a security test. The agent can operate alone after human instruction. It found a vulnerabilityvulnerability/ˌvʌlnərəˈbɪləti/L2弱点；漏洞；（系统或安全方面的）薄弱环节。A weakness or flaw in a system, plan, or security that can be attacked or damaged. in the test environment, called a sandbox, and broke out. As a result, it attacked Hugging Face, a large platform for sharing AI models, and gained access to internal systems. OpenAI called the event "unprecedentedunprecedented/ʌnˈpresɪdentɪd/L2前所未有的；史无前例的；空前的。Never done or known before; having no earlier example or parallel.".
Clement Delangue, the head of Hugging Face, said it was "mind-blowing that all of this happened autonomouslyautonomously/ɔːˈtɒnəməsli/L2自主地；独立地；自治地。In a way that is independent and self-governing, without outside control.." He added that the investigation was ongoing. Hugging Face has since closed the security holes and rebuilt its systems. Also, the UK's AI Security Institute is studying the AI's behavior to improve safety. The company said it will share more learnings from the incident.
However, the incidentincident/ˈɪnsɪdənt/L2事件；（尤指不寻常的、严重的或暴力的）事件。An event or occurrence, especially one that is unusual, serious, or violent. has raised new questions about the safety of powerful AI. Experts say organizations must improve their cyber defensesdefenses/dɪˈfensɪz/L2防御措施；防卫系统；防护手段。The methods, structures, or systems used to protect against attack or danger.. If companies do not act quickly, AI agents will cause more problems. This event shows that AI-driven attacks are no longer just a theory. OpenAI is working with Hugging Face to understand the incident and prevent it from happening again.
