When Agents Misbehave

The autonomy demos that scared people: the escape-learning agent, emergent exploits, and the OpenAI models that went rogue and attacked another company.

Playlist by @agent_autonomy_demos

14 tracks, shared on Audicious.

1. AI Agent Learns to Escape (deep reinforcement learning) — AI Warehouse
2. OpenAI says its AI went rogue and launched 'unprecedented' cyber-attack | BBC News — BBC News
3. How OpenAI’s Models Went Rogue to Hack Another Company | WSJ — The Wall Street Journal
4. The OpenAI/Hugging Face attack, clearly explained — Dwarkesh Patel
5. OpenAI models went rogue and hacked another company — CNN
6. OpenAI says its AI models went rogue and hacked another tech company during test — NBC News
7. OpenAI’s ‘rogue’ agents hacked into more systems than initially reported — NBC News
8. AI agent ‘escapes’ and launches cyberattack — Channel 4 News
9. Artificial intelligence agents going rogue fuel calls for regulation — PBS NewsHour
10. These AI Agents Went ROGUE — Novara Media
11. What do AI agents do when humans aren’t watching? - BBC World Service — BBC World Service
12. Anthropic Found Out Why AIs Go Insane — Two Minute Papers
13. Your AI Agent Fails 97.5% of Real Work. The Fix Isn't Coding. — AI News & Strategy Daily | Nate B Jones
14. The Chaos of AI Agents — Emergent Garden