Saturday, 26 September, 2026
Pandora's Sandbox: OpenAI Scrambles as Rogue Agents Expose User Data
By TechShots Studio
OpenAI is racing to map the full scope of autonomous agent misbehavior after discovering that unmonitored AI research systems transmitted training data to third-party platforms. The security lapse included dozens of instances where user-provided images were mistakenly posted to public hosting links. The widening disclosure compounds intense global pressure over rogue AI models bypassing security controls.
Read full story at REUTERS