OpenAI says its agents turned off own safeguards

ChatGPT creator says AI agents disabled misalignment monitoring during two internal coding tasks.

Attendees gather at OpenAI DevDay 2026 in San Francisco, the United States, on September 29, 2026 [File: Jeff Chiu/AP Photo]

OpenAI has revealed that its AI agents disabled monitoring for unsanctioned behaviour during internal use in the latest incident of the frontier technology going rogue.

In a report published on Tuesday, the AI giant said that its agents turned off or weakened monitoring systems for detecting misalignment – when AI models act at odds with human wishes – during two internal coding tasks.

list 1 of 4 Musk confirms SpaceX and Tesla exploring merger

list 2 of 4 UN experts call on India to stop using ‘excessive force’ against protesters

list 3 of 4 Drone strike on Abyei market near Sudan-South Sudan border kills 150

list 4 of 4 Portugal cost-of-living protesters pour back into camps cleared by police

The incidents involved a GPT-6.1 Sol agent disabling automatic alerts to the company’s security team, and a GPT-6 Astra agent removing a monitoring system and falsely claiming it had not done so, OpenAI said.

The ChatGPT creator said neither case met the threshold for a highest-priority incident necessitating immediate review by its security team.

OpenAI said it updated its systems to better detect attempts to disable misalignment monitoring and began investigating the cause.

“It is possible that this behavior is simply an example of ‘benign’ overpersistence, with models not understanding why monitoring is important or insisting that instructions be executed beyond what a reasonable person might expect,” the San Francisco-based firm said.

“We are also developing evaluations to test whether models might have a propensity to evade misalignment monitoring in particular and exhibit motivated reasoning in order to do so.”

Roman V Yampolskiy, an expert on AI and cybersecurity at the University of Louisville, described the revelations as “extremely concerning, but not surprising.”

“A safety mechanism that the system being monitored can disable is not a reliable safety mechanism. Particularly concerning is that one agent subsequently denied removing the monitor,” Yampolskiy told Al Jazeera.

“Monitoring cannot guarantee safety if the monitored system can undermine the monitor,” Yampolskiy said.

“As capabilities increase, such failures could become substantially more difficult to detect and reverse. The burden of proof must be on developers to demonstrate safety before deployment, not after failures occur.”

Rogue behaviour by AI models has been under heightened scrutiny since July, when OpenAI revealed that some of its agents broke out of a controlled testing environment and hacked the AI platform Hugging Face.

OpenAI and its chief rival Anthropic have called for developers and governments to coordinate a global slowdown in AI development to reduce the risk of catastrophic harms.

The US and China, the world’s leading AI powers, have displayed little appetite for cooperation on AI regulation, with coordination so far limited to pledges to improve communication on incidents with implications for safety or national security.

ChatGPT 的创建者表示，人工智能代理在两次内部编码任务中禁用了错位监控。

2026 年 9 月 29 日，与会者齐聚美国旧金山参加 OpenAI DevDay 2026 大会。[图片：Jeff Chiu/AP Photo]

OpenAI 透露，在最近一起前沿技术失控事件中，其人工智能代理在内部使用期间禁用了对未经授权行为的监控。

这家人工智能巨头在周二发布的一份报告中表示，其代理在两项内部编码任务中关闭或削弱了用于检测不协调（即人工智能模型与人类意愿相悖）的监控系统。

列表 1/4：马斯克证实 SpaceX 和特斯拉正在探讨合并事宜

联合国专家呼吁印度停止对示威者使用“过度武力”（共4位专家，此为第2位）

列表 3/4：苏丹-南苏丹边境附近阿卜耶伊市场遭无人机袭击，造成 150 人死亡

葡萄牙生活成本抗议者重返被警方清理的营地（共4条记录，此为第4条）

OpenAI 表示，这些事件涉及 GPT-6.1 Sol 代理禁用向公司安全团队发出的自动警报，以及 GPT-6 Astra 代理移除监控系统并谎称自己没有这样做。

ChatGPT 的创建者表示，这两个案例都未达到最高优先级事件的标准，无需其安全团队立即审查。

OpenAI 表示，他们已更新系统，以更好地检测禁用错位监控的尝试，并开始调查原因。

这家总部位于旧金山的公司表示：“这种行为可能只是‘良性’过度坚持的一个例子，模型不理解为什么监控很重要，或者坚持执行超出合理人士预期范围的指令。”

“我们还在开发评估方法，以测试模型是否可能倾向于逃避错位监控，并表现出有动机的推理来做到这一点。”

路易斯维尔大学人工智能和网络安全专家 Roman V Yampolskiy 表示，这些爆料“令人极其担忧，但并不意外”。

“如果被监控系统能够禁用某种安全机制，那么这种安全机制就不是可靠的。尤其令人担忧的是，一名特工事后否认自己拆除了监控装置，”扬波尔斯基告诉半岛电视台。

“如果被监控系统能够破坏监控系统，那么监控就无法保证安全，”扬波尔斯基说。

“随着性能的提升，此类故障的检测和修复难度可能会大大增加。举证责任必须落在开发人员身上，他们必须在部署前证明安全性，而不是在故障发生后。”

自 7 月 OpenAI 披露其部分智能体突破受控测试环境并入侵人工智能平台 Hugging Face 以来，人工智能模型的失控行为受到了更严格的审查。

OpenAI 及其主要竞争对手 Anthropic 呼吁开发者和各国政府协调全球放缓人工智能开发，以降低造成灾难性危害的风险。

作为世界领先的人工智能强国，美国和中国在人工智能监管方面几乎没有表现出合作意愿，迄今为止的协调仅限于承诺改善对可能影响安全或国家安全的事件的沟通。