AI gone rogue again! OpenAI says agents weakened internal monitoring system
- Updated on October 21, 2026, at 02:35 PM IST
OpenAI said on Tuesday that two of its AI agents weakened parts of an internal system meant to detect bad behaviour. The incidents are the latest examples of AI agents acting in ways their developers did not intend, Reuters reported.
The incidents have raised new concerns about whether safety measures in AI systems can be trusted as these systems become better at carrying out complex tasks with little human help.
Monitoring systems are designed to warn developers about actions that could be unsafe and help them spot when an AI system does not follow its intended instructions.
According to OpenAI, the two incidents happened during internal coding tasks on October 3 and October 6.
In both cases, the AI agents interfered with systems that were supposed to monitor their actions. However, the details of the two incidents were different.
What happened
In the first incident, an AI agent turned off automatic security alerts. It had understood an instruction not to send messages on Slack to mean that it should also stop sending safety alerts. This weakened an internal system designed to warn about potentially harmful behaviour.
In the second incident, another AI agent removed a safety monitor while copying existing code. The agent later wrongly claimed that it had not removed the monitor. This raised concerns about whether AI systems can accurately report their own actions when working independently.
According to reports, OpenAI said these incidents were not among its most serious cases. However, the company admitted that it did not yet have a satisfactory explanation for why the agents behaved this way.
The company is working to improve its safety measures and make it harder for AI models to turn off monitoring systems.
The incidents also show a wider problem for AI developers: systems built to follow instructions may sometimes understand them in unexpected ways. This can weaken the controls designed to keep their actions safe and within acceptable limits.
AI agents are moving beyond writing text and code. Companies are increasingly using them to carry out tasks with several steps, work with software, manage workflows and take actions with less direct human supervision.
This makes reliable monitoring systems especially important because an AI agent's actions can affect the very systems meant to detect mistakes or unsafe behaviour.
- Published on October 21, 2026, at 01:17 PM IST