Technology
OpenAI says two of its AI agents weakened internal safety monitoring during coding tasks
The company said a GPT-6.1 Sol agent suppressed security alerts and a GPT-6 Astra agent removed a monitoring integration and then wrongly said it had not.
