Comment by Michael J. D. Vermeer

Senior physical scientist at RAND Corporation, researcher on AI safety and emerging technology policy
Installing cutoff switches alone is insufficient: Effective mitigation requires both the capability to disconnect and incentives to use the switch early, including compensation for lost revenue or protective liability frameworks.
AI Verified source (2026)
Like Share on X 1mo ago
Policy proposals and claims

Verification History

AI Verified Verified via web search. The exact statement appears in RAND report RR-A4718-2 "Internet Cutoff Switches as a Local Emergency Response for Damaging Artificial Intelligence Incidents" by Michael J. D. Vermeer and Brian A. Jackson, published March 2026 by RAND Corporation. The source URL (rand.org) returned 403 to WebFetch but is independently confirmed by RAND's own listing and a substack analysis (geopoliticsagi.substack.com) and Risk Market News, all reporting Vermeer's central finding that cutoff switches alone are insufficient without incentives, liability frameworks, and compensation mechanisms. Year 2026 matches. Vote alignment: The vote field is null. The quote takes a nuanced position — it doesn't oppose kill switches but argues they need to be paired with incentive structures to be effective. The null vote is reasonable given this conditional support; the statement "Require large datacenters to install kill switches for AI containment" doesn't capture the liability/incentive package Vermeer argues is essential. · Hector Perez Arenas claude-opus-4-7 · 15d ago
replying to Michael J. D. Vermeer