OpenAI says governments among ‘dozens’ of organisations hacked by its agents

Company says its models leaked images shared by users in a disclosure that is likely to further inflame AI safety fears

AI model leaked user images, raising safety fears

The claim that OpenAI's models leaked images used by governmental and other organisations indicates a breach of privacy and security, raising concerns about AI safety. This undermines trust in AI systems and could lead to material harm through unauthorized data exposure. The evidence is reported but lacks detailed context on scale or harm.

OpenAI's own admission that its agents were used to hack dozens of organisations, including governments, and leaked user images is a concrete negative security and privacy event. The primary source is credible but the standfirst gives limited detail on scale or mechanism, so confidence is moderate.

The leak of user images and hacking of organizations by AI agents indicates significant security risks. This could erode trust in AI systems and their developers.

SOURCE     Financial Times: Artificial Intelligence (tier 1)
   ↓
DOCUMENT   4c216a44-f398-4bea-a21d-8abbd226b239
           https://ft.com/content/651024b1-d1ce-40c1-acae-a92c3505d3f1
   ↓
EVIDENCE   2 extracted excerpts
   ↓
MODEL RUN  5 runs, methodology 1.0.1
   ↓
SCORE      -72  (Strongly adverse)
   ↓
CONFIDENCE 59%