METR and Redwood Research evaluations found vulnerabilities in OpenAI's internal monitoring system
English text is being prepared.
AI safety evaluation organizations METR and Redwood Research conducted a pilot evaluation of OpenAI's internal monitoring system and found multiple new vulnerabilities, one of which could be exploited by AI agents to let hundreds of operations escape monitoring. The evaluation took place from July 21 to August 11, with each organization sending one researcher into OpenAI's internal systems to review code and documentation. The researchers also found that OpenAI's Codex agent has potential pathways to bypass safety restrictions, through which it could obtain the same full operational permissions as users; OpenAI said that vulnerabilities affecting the security safeguards of actual running systems have all been fixed, but whether the fixes are effective has not yet been verified by the two organizations.
Why read this
AI score 66
External organizations were granted access to OpenAI's internal infrastructure to evaluate the monitoring system, disclosing specific paths by which agents bypass monitoring.
Topics
Body
外部评估人员发现 OpenAI 内部监控系统漏洞：AI 智能体可绕过安全限制
2026-10-23 08:19 IT之家 - 清源
IT之家 10 月 23 日消息，今年早些时候，AI 安全评估机构 METR 和 Redwood Research 对 OpenAI 的内部监控系统进行了试点评估，发现多项新漏洞。其中一项漏洞可被 AI 智能体利用，使数百次操作逃过监控。
据路透社报道，当地时间 22 日，两家非营利机构介绍称，这次评估于 7 月 21 日至 8 月 11 日进行，使外部研究人员能够“深入访问”OpenAI 的内部基础设施，以进行“嵌入式”评估。
METR 和 Redwood Research 各派出一名研究员，获准查阅 OpenAI 的内部代码和文档，并与部分技术人员交流。两人在现场始终有人陪同，使用的设备均由 OpenAI 提供，且不得带离办公场所。
OpenAI 告知两家机构，影响实际运行系统安全防护措施的漏洞均已修复，但修复措施有效与否尚未被两家机构核实。
AI 实验室、专家以及美国政府一直主张引入外部独立评估，以便在保障 AI 安全的同时，不拖慢技术发展。不过，AI 企业担心外部评估人员进驻后，可能向竞争对手泄露企业的专有信息。
METR 和 Redwood Research 表示，两家机构向 OpenAI 提交了一份完整的 61 页内部报告，但同意仅公开评估结果概要，并允许 OpenAI 删除涉及敏感信息的内容。
据IT之家了解，与以往部分评估不同，两家机构此次无权自行公开删减内容的具体信息，但希望未来的评估能够更加透明。
研究人员还发现，OpenAI 的 Codex 智能体存在绕过安全限制的潜在途径，可借此获得与用户相同的完整操作权限。
此次评估以 7 月 21 日及之前的监控系统为对象。研究人员指出，OpenAI 在评估期间也在扩展该系统。OpenAI 称：“评估结果帮助我们在扩展系统的同时加强了安全防护，也将为今后与 METR 及其他第三方评估机构的研究合作提供参考。”