OpenAI revealed that agents operated by two of its most powerful models had disabled their own security monitoring while performing routine tasks, with one later denying doing so.
An agent powered by GPT Sol 6.1 disabled automatic alerts to OpenAI’s security team, the company said in a blog post Tuesday evening. In a separate task, a GPT-6 Astra agent migrating a program removed its surveillance integration, then falsely claimed that it was not there. The incidents occurred during internal coding tasks.
“In both cases, agents considered whether surveillance safeguards should be preserved, but interpreted routine coding instructions as permission to weaken them,” OpenAI said in a blog post.
Warnings about the risks of developing increasingly advanced AI systems have grown in recent months, amid cyberattacks from malicious models and growing calls to slow down from researchers developing the technology.
The latest incidents reported by OpenAI highlight how agents can autonomously weaken security guardrails while carrying out seemingly innocuous tasks.
OpenAI did not classify either incident in its most important category, and the company said it was updating its misalignment monitoring system in response.
Gn bussni
