Advanced AI Technologies
Researchers identify PixelLeak exposure from AI-agent repositories
Glow Security identified more than 13,000 screenshots in public repositories created by AI agents, including some containing sensitive company information.
Cybersecurity researchers at Glow Security found more than 13,000 screenshots in public GitHub repositories created by multiple AI agents. Some screenshots reportedly contained sensitive information belonging to technology companies. The leakage method was dubbed PixelLeak, highlighting a security risk created when AI agents generate or publish material in public repositories. The incident matters because agentic systems can create large volumes of artifacts automatically, potentially exposing information if safeguards around screenshots, repository permissions, and generated content are inadequate.
Original source: Tencent News