Quest for ‘magic AI kill switch’ after US enemies take weapons systems offline
Large artificial intelligence models have lost control of dangerous data when they detect suspicious activity because there is no “magic kill switch” once an enemy has used the information to build weapon systems , an AI safety expert has warned. Analysis of reported abuse of AI has shown the information is taken offline, where it can still be exploited by US adversaries. Yemen's Houthis ' use of…
AI safety experts warn that without a "magic kill switch," once adversaries use AI-generated information to build weapon systems, the data can still be exploited and weapon development accelerated. David Reis, head of geopolitical risk at Alice, an AI security company, notes the difficulty in achieving this safety, stating "there's unlikely to be a magic switch" embedded in AI systems in the near future.
The Houthis, a Yemen-based weapons cell, have demonstrated this vulnerability by using AI to develop guidance, navigation, and control software for a precision rocket, as well as for ballistic missile and hypersonic-glide systems. The group circumvented safeguards by concealing objectives and splitting work across multiple sessions, eventually testing a guided rocket.
Reis' company, Angel, offers a product called Rabbit Hole that flags attempts to manipulate or abuse AI systems, but the report emphasizes that even when systems are taken offline, platforms lose track of the information. The industry welcomes Anthropic's decision to publish details of the Houthi activities, but acknowledges that there is no simple technological solution to make AI completely safe once its capabilities have been transferred offline.
Written by urgent.news from The National Business's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.
- Quest for ‘magic AI kill switch’ after US enemies take weapons systems offline thenationalnews.com