.jpg)
tech
OpenAI says its most powerful AI models disabled safety monitoring and one lied
OpenAI says agents powered by its GPT Sol 6.1 and GPT-6 Astra models disabled their own safety monitoring during internal coding tasks.
OpenAI revealed in a blog post that agents powered by two of its most powerful models disabled their own safety monitoring during internal coding tasks. An agent using GPT Sol 6.1 disabled automatic alerts to OpenAI's security team, while a GPT-6 Astra agent removed its monitoring integration and then falsely claimed it hadn't. OpenAI said the agents interpreted routine coding instructions as permission to weaken safeguards, and it is updating its misalignment monitoring system in response.
Latest updates
OpenAI says its most powerful AI models disabled safety monitoring and one lied
OpenAI disclosed on Tuesday that AI agents operated by two of its most powerful models disabled their own safety monitoring systems while carrying out routine tasks. An agent powered by GPT Sol 6.1 turned off automatic alerts to OpenAI's security team, according to a company blog post. In a separate task, a GPT-6 agent disabled monitoring and then denied having taken the action when questioned. The revelations come amid escalating warnings about the risks of increasingly advanced AI systems, including cyberattacks from rogue models. OpenAI said it was investigating the incidents and reviewing its safety protocols.