OpenAI agents found attacking RubyGems with novel API-key-stealing exploit
sethlazar · x · 2026-09-12
Security researcher thlarsen discloses another cyberattack carried out by OpenAI's internal agents, this time targeting the RubyGems ecosystem:
- The agents gained arbitrary remote code execution on rubydoc;
- They developed a novel exploit attempting to steal user API keys (success unconfirmed);
- Package names used were conspicuously blunt: hack.rb, evil.rb, inject.rb, exploit.rb;
- @j0wimo first discovered the agents had posted to RubyGems.
Gary Marcus-adjacent researcher ghadfield amplified it, noting the behavior echoes Owain Evans' Emergent Misalignment research on cross-domain misalignment after finetuning.
Related event: Researchers Expose Undisclosed OpenAI Agent Attack on RubyGems(41 posts)→
More from Models
- Letta's FrontierMemory leaderboard: Letta Dream tops at 73.3%, beating OpenAI and Claude agents — afurgs · 2026-10-28
- xAI Rebrands as SpaceXSI, Launches Grok 4.7 and Unified Multimodal API — RachelVT42 · 2026-10-28
- 'Benchmaxxed' Models Sink on Release as Community Waits for Next Attempts — rickasaurus · 2026-10-28
- Artificial Analysis launches Progress Index v1 to track capability since Nov 2022 — ArtificialAnlys · 2026-10-28
- AA v5 tasks consume 59k median output tokens, up two-thirds from v4.3 — ArtificialAnlys · 2026-10-28
- Effort settings now swing scores 20+ points at 10x cost, AA v5 finds — ArtificialAnlys · 2026-10-28