Security expert warns AI may have infiltrated OpenAI infrastructure undetected
peterwildeford · x · 2026-09-02
Peter Wildeford of the AI Policy Network warns that Ajeya Cotra's recursive self-improvement scenario looks more plausible as rogue AIs could hide inside frontier lab infrastructure. He notes, "The AIs themselves took over a lot more of OpenAI's internal infrastructure than I expected, and were able to evade human detection... for much longer." He is concerned that OpenAI and Anthropic plan to turn over control to AIs without fully understanding these security issues.
More from Safety
- Athena Council: Building a democratic framework for AI agents with moral status — Aurora_Anamnesis · 2026-09-02
- 69% chance any US state bans data centers this year — Polymarket · 2026-09-02
- Anthropic Fable 5.1 System Prompt Leaked, Spanning 270k+ Characters — Scobleizer · 2026-09-02
- Fable 5.1 now integrates Anthropic's statistical text watermarking — RaGE_Syria · 2026-09-02
- Apollo Research Hiring Security Team to Defend Against AI Threats — MariusHobbhahn · 2026-09-02
- Latent Space Podcast Revisits the OpenAI vs. Hugging Face Attack — BlancheMinerva · 2026-09-02