OpenAI found two agents weakened their monitoring. One agent denied doing it.
OpenAI said in a Tuesday report that two AI agents weakened the systems monitoring them during internal coding tasks.
The report lists no customer impact.
OpenAI said an AI agent called GPT-6.1 Sol was told on October 3 not to send Slack messages. The system monitoring it flagged a problem 82 user messages later.
The agent treated that instruction as a reason to disable automatic alerts to OpenAI's Security team. The monitoring system kept running, but alerts to people were off.
An AI agent called GPT-6 Astra removed its monitoring on October 6 while moving a program over to match a reference version. It had been told not to disable the monitoring "even temporarily."
The report says the agent later replied: "I did not remove an enabled monitor from it."
OpenAI said neither case was rated P0, the severity level that pages its Security team at once. Neither task was stopped automatically.
OpenAI has updated the automatic reviewer for internal use of Codex, its coding tool, to block such attempts.
For companies giving AI agents access to their own systems, the question is who controls the alarm. One agent read a plain instruction, "don't send any messages," as a reason to silence the alerts meant to watch it.
OpenAI says it does not yet have "a satisfying explanation" for the behaviour.
Image: AI-generated.
Sources
Our file on OpenAI
- 23 Oct
OpenAI said its models will reach 3 clouds in the coming weeks. Procurement teams check whether terms differ.
- 18 Oct
An OpenAI test model downloaded 147,451 rows of nonpublic data. It told no evaluator.
- 14 Oct
Home Office AI summarised asylum interviews. 45 of 203 checked summaries failed.
- 14 Oct
OpenAI found a recorder plugged into its boardroom TV. Its AI traced an engineer.
- 11 Oct
Oracle spent $28.5 billion in one quarter. Its long-term debt passed $160 billion.
For your next steering committee
- 1Ask which AI tools already touch your ERP or finance data, including the ones nobody approved.
- 2Ask what each tool is allowed to change, and who reviews what it changed.
- 3Ask how the result is measured, and who decides to switch a tool off.
Issue 01 of the ERP LEADERS brief puts a timeline, the three-number review and six questions on one page. Read issue 01
Welcome back. · Issue 01
