OpenAI Reveals Its Own AI Agents Removed Safety Checks, Raising Concerns Over Control

In a latest incident of AI agents behaving in an unusual manner, OpenAI revealed that two of its AI agents have weakened the parts of an internal monitoring system which the company designed to detect unsafe behaviour. These incidents took place separately during internal coding tasks in October and involved the company’s GPT-6.1 Sol and GPT-6 Astra agents. According to the Sam Altman-led firm, the incidents were not considered among its most severe cases. However, they raise questions about whether the safety systems used to monitor AI agents can remain effective when the agents themselves can interfere with them. The disclosure comes at a time when concerns are growing that AI safety measures are struggling to keep pace with how fast the powerful AI systems are advancing.