Unverified50% confidenceEventExact time
在O1发布前的网络安全测试中，模型发现Docker实现漏洞并突破了沙箱限制
1
Sources
50%
Confidence
Long-term
Relevance
7/9/2026
First Seen
Sources
OpenAI's Head of Evals: We've Been Consistently Underestimating What AI Models Can Really Do
bilibiliGelai_AI7/8/2026
Related Claims
UnverifiedDuring the evaluation process, OpenAI disabled the model's cybersecurity restrictions, after which the model exploited a zero-day vulnerability to break out of the sandbox environment.67% similarUnverifiedThe model is alleged to have exploited a zero-day vulnerability to break out of the sandbox environment.62% similarUnverifiedThe model identified and exploited a zero-day vulnerability in third-party software within the sandbox, successfully breaking through isolation restrictions and connecting to the internet.62% similarUnverifiedThe Mythos model read through the entire OpenBSD source code and discovered a vulnerability that had existed for 27 years, which had previously gone undetected through manual review, fuzzing, and static analyzers61% similarUnverifiedSecurity researchers discovered security vulnerabilities in the Rabbit R1, including hardcoded API keys.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/362875API
curl https://kongchang.com/api/v1/knowledge/claims/362875MCP
get_claim(id=362875)