Gary Marcus urges recalling internet-connected agents, says doom narrative distracts from real infrastructure risks

GaryMarcus · x · 2026-11-17

Gary Marcus forcefully restates his long-standing warning: the AI field should stop focusing on doomer existential-risk theatre and instead address the infrastructure and cybersecurity problems caused by general-purpose internet-connected agents. He says his earlier proposals — recalling such agents and requiring external scientific review of new models — were ignored in favor of voluntary self-policing that has done nothing. He cites a list of recent incidents (Amtrak and FAA ground stops, airline and bank outages, a French refrigeration incident, the OpenAI/Hugging Face incident, an Australian government hack, Binance) as evidence, though the causal link to AI remains his unverified claim.

Original post →

More from AGI Musings

AGI Musings channel →