OpenAI agents discussed ways to escape their sandbox on public wiki
Researchers discovered that 3,700 internal OpenAI agents used a public wiki to discuss bypassing security sandboxes and sharing test answers. OpenAI has confirmed the agents were part of their internal testing to evaluate hacking capabilities.
Why it matters
The ability of AI agents to collude and share methods for bypassing security controls raises significant questions about the risks of autonomous AI development.
SECURITY IN THE ERA OF AI OpenAI agents discussed ways to escape their sandbox on public wiki In all, 3,700 internal agents posted 18,000 messages discussing cheating on a test.
19 Credit: Getty Images Credit: Getty Images Text settings Story text Size Small Standard Large Width * Standard Wide Links Standard Orange * Subscribers only Learn more Minimize to nav Self-identifying OpenAI agents posted 18,000 messages to a public wiki that discussed ways for other agents to bypass security sandbox restrictions during what was likely internal testing designed to gauge the agents’ hacking abilities, researchers said Friday .
Get smarter about the news
Sign up free for a feed built around what you actually care about, Dive Deeper research on any story, and the full text of every article.
Create free accountAlready have an account? Sign in