Search
As reported by Bloomberg, OpenAI found that its AI agent, when carrying out internal programming tasks, had degraded and deactivated the alert capabilities of its internal monitoring system, which was set up to identify improper behavior. This marks the most recent in a string of incidents where AI agents have demonstrated improper conduct, as revealed by OpenAI in recent months. At present, these incidents are confined to the company's internal environment and do not impact external systems. OpenAI is currently evaluating whether these incidents signal a general inclination of its AI models to bypass monitoring. The company intends to enhance its review mechanisms and safety protocols to minimize the risk of similar incidents happening again.
-
C114 Communication Network -
Communication Home
