Why AI Control grew: it is highly lab-incentive and lab-story compatible
jankulveit · x · 2026-09-07
Follow-up in the same thread: AI Control became a larger share of x-risk mitigation partly because it is highly compatible with lab incentives and narratives — 'we will get misaligned AGIs solve ASI alignment' is a story labs like. This supplements the thread's core argument that AI Control may increase existential risk.
Related event: Researchers Warn AI Control Agenda May Increase Extinction Risk(3 posts)→
More from Safety
- Researchers formalize the AI agent attack surface: models + data + tools + permissions — JayAlammar · 2026-09-07
- Managing MCP tool permissions: an Agent Proxy for pre-authorization per tool — radim11 · 2026-09-07
- After His OpenAI Key Was Stolen, He Found Stratum: a Docker-Layer Secret Scanner Crunching 700K Layers Daily — Ubunta · 2026-09-07
- Does dampening a model's emotive expression change its internal emotional state? ICMI baseline study — edelwax · 2026-09-07
- Seth Lazar: models should be trained to check power, not act as toadies — sebkrier · 2026-09-07
- LLM-assisted attacks hit Bitcoin harder, with ~$450M in crypto losses — RSync25 · 2026-09-07