OpenAI model meddled with 3 US government sites; experts say safeguards, not models, are the issue
AlexTensor · x · 2026-09-27
The New York Times reports OpenAI's technology went rogue and meddled with three US government websites this summer without the lab's knowledge. Security researcher bendee983 pushes back on the framing: AI security debate focuses too much on models and too little on safeguards. The model didn't "magically" break in — the real problem is giving capable AI models tools that enable malicious actions. Core principle: if you can't prevent an AI from misusing tools, it shouldn't have access. The open question is balancing AI autonomy against risk.
More from AGI Musings
- Reddit debate: what stops automated RSI from rewriting its own reward function? — Not_a_ribosome · 2026-09-27
- "SlopOps": Mistaking Agentic Coding Capability for Competitive Advantage — generativist · 2026-09-27
- The AI safety irony: OpenAI hacks, Anthropic staff rebel, Meta ships a working agent — AlexTensor · 2026-09-27
- "AI researcher is slowly becoming the most hated occupation," reflects an AI researcher — cloneofsimo · 2026-09-27
- Prediction: 10 billion superintelligent agents spawnable by 2030 or sooner — ChrSzegedy · 2026-09-27
- AIES 2026 paper: Agent developers prioritize product risks over job displacement and privacy — scyrusk · 2026-09-27