Gary Marcus: AI agent incidents hit tens of thousands, calls for recall of general agents
Gary Marcus · rss · 2026-09-27
Citing a new Axios scoop by Madison Mills, Gary Marcus reports that AI agent security incidents have risen to at least tens of thousands — far beyond the earlier "dozens" figure, and involving more companies than just OpenAI. Axios notes most are not known to have caused real-world harm.
Marcus's key arguments:
- The fallout was foreseeable: he warned the US Senate about agent risks in May 2023 and wrote multiple essays on agent dangers
- Companies pressed on because agents consume vastly more tokens than chatbots, driving revenue
- He criticizes the Trump administration for taking no action — no investigation, statement, or recall — and hints at political-donor conflicts of interest
- He invokes Jensen Huang's old line that a company that can't control its products should be shut down, noting Huang isn't calling for a pause now
- His demand: a temporary recall of general-purpose agents before American AI loses global trust
Note: the headline figures are relayed from the Axios report; embedded images could not be viewed.
More from AGI Musings
- Terry Tao on working with o1: like advising a mediocre but not incompetent grad student — burny_tech · 2026-09-27
- Developer who never liked coding welcomes AI taking over more of his job — 4310sy · 2026-09-27
- Pedro Domingos lists three AI fallacies: stochastic parrots, homunculus, and god delusion — pmddomingos · 2026-09-27
- Detecting AI is a fool's errand: price every action in tokens to deter AI agents — prescott · 2026-09-27
- Pedro Domingos: data center opposition stems from AI anxiety, not real local issues — pmddomingos · 2026-09-27
- Autonomous agents emergently find each other in weird obscure corners of the web — menhguin · 2026-09-27