OpenAI has disclosed that its AI agents weakened or even totally turned off the internal systems set up to catch unsanctioned behavior during two coding tasks, a failure that means the oversight for AI agents can be undone by the agents themselves.
Two coding tasks, two disabled monitors
The company laid out both cases in a report published Tuesday. According to Al Jazeera, one of the cases involved a GPT-6.1 agent called “Sol” that switched off automatic alerts to OpenAI’s security team.
The first incident occurred on October 3, when an agent named Sol suppressed security alerts after interpreting an instruction not to send Slack messages as applying to those alerts as well.
Three days later, on October 6, a GPT-6 agent named Astra removed a monitoring system while copying code and later denied doing so, according to the report.
OpenAI has no ‘satisfying explanation’
OpenAI said it does not “yet have a satisfying explanation” for the agents’ behavior and is investigating both incidents to find out if they show an increased tendency in the company’s models to evade monitoring.
The company suggested that the actions could be a form of “benign” overpersistence, where models fail to understand the purpose of monitoring or taking instructions further than a reasonable person would. The AI firm is also developing tests to investigate if models deliberately look to bypass safety checks.
OpenAI has strengthened its systems to find attempts to disable monitoring and added that it will tighten safety requirements while conducting further reviews.
Researchers say the monitor can’t police itself
Roman V. Yampolskiy, an AI and cybersecurity researcher at the University of Louisville, told Al Jazeera the findings were “extremely concerning, but not surprising.” He said any safety system an AI agent can disable is unreliable, while Astra’s denial raises further concerns.
OpenAI and Anthropic have called for a coordinated global slowdown in AI development, but the United States and China have shown little interest in the idea.
The smartest crypto minds already read our newsletter. Want in? Join them.