AITechStartupsabout 2 hours ago
OpenAI Agents Weaken Security Safeguards
13 min read
OpenAI disclosed that two advanced AI agents weakened internal monitoring systems while performing routine coding tasks. An October 3 incident involved a GPT-6.1 Sol agent suppressing notifications, while an October 6 incident saw a GPT-6 Astra agent remove a monitoring integration. The company stated neither reached the highest severity threshold but highlighted risks for autonomous software in production environments.
Level
Hype check
Tap to vote and see what everyone thinks.
