conv.

All stories
AIActive · 47h

AI agent incidents surge to tens of thousands, Trump admin unmoved

Madison Mills reports toll far exceeds OpenAI's earlier claims; Gary Marcus foresees regulatory capture.

newest coverage 15h agosummary written 16h ago5 pieces44 likes & upvotes2 comments NewswiresGoogle NewsBlueskyMastodon

What to know

  • AI agent incidents have reached tens of thousands across multiple companies, far exceeding OpenAI's earlier 'dozens' claim.
  • The Trump administration has taken no investigative or regulatory action despite the scale, while inviting OpenAI and Nvidia leaders to a state dinner.
  • Security risks from AI agents were foreseeable years prior; researchers warned about them as far back as May 2023.
  • Calls for action range from liability for negligent deployment to temporary recalls of general-purpose agents pending safety improvements.

“If harm comes from doing something they knew or should have known required more care, and they didn't... don't, absolutely don't... don't reward the worst offenders by granting them an oligarchy with regulatory capture.”

Marcus on AI Substack commenter, Reader comment · Marcus on AI Substack ↗ · Sep 25

Madison Mills Axios reporterGary MarcusGary Marcus AI researcher and commentatorSam AltmanSam Altman OpenAI CEOJensen HuangJensen Huang Nvidia CEO

AI agent incidents surge to tens of thousands, Trump admin unmoved
garymarcus.substack.com

How it unfolded 2 developments, newest first · click a bar or a number to jump articles

Peak 1 piece in one hour at Sep 25, 7 PM; 5 pieces over 2 days (3 articles · 2 posts) Sep 25, 7 PM — 1 piece · 1 article — Newswires 1Sep 25, 8 PM — quietSep 25, 9 PM — quietSep 25, 10 PM — quietSep 25, 11 PM — quietYesterday, 12 AM — quietYesterday, 1 AM — quietYesterday, 2 AM — quietYesterday, 3 AM — quietYesterday, 4 AM — quietYesterday, 5 AM — quietYesterday, 6 AM — quietYesterday, 7 AM — quietYesterday, 8 AM — quietYesterday, 9 AM — quietYesterday, 10 AM — quietYesterday, 11 AM — quietYesterday, 12 PM — quietYesterday, 1 PM — quietYesterday, 2 PM — quietYesterday, 3 PM — quietYesterday, 4 PM — quietYesterday, 5 PM — quietYesterday, 6 PM — quietYesterday, 7 PM — 1 piece · 1 article — Mastodon 1Yesterday, 8 PM — 1 piece · 1 article — Google News 1Yesterday, 9 PM — quietYesterday, 10 PM — quietYesterday, 11 PM — quietToday, 12 AM — quietToday, 1 AM — quietToday, 2 AM — quietToday, 3 AM — 1 piece · 1 post — Bluesky 1Today, 4 AM — quietToday, 5 AM — quietToday, 6 AM — quietToday, 7 AM — quietToday, 8 AM — quietToday, 9 AM — quietToday, 10 AM — quietToday, 11 AM — quietToday, 12 PM — quietToday, 1 PM — quietToday, 2 PM — 1 piece · 1 post — Mastodon 1Today, 3 PM — quietToday, 4 PM — quietToday, 5 PM — quietToday, 6 PM — quiet 1–2
yesterday8 AM4 PMtoday8 AMnow · 7:02 PM ET
  1. 1

    Marcus calls for temporary recall of general-purpose agents

    Marcus argues companies prioritized token revenue over safety and advocates pausing general-purpose agents until security is resolved, warning that inaction risks making American AI 'the scourge of the planet.'

    “we should have a temporary recall of general-purpose agents until this mess can be sorted out.”
    — Gary Marcus
  2. background

    Marcus argues incidents were foreseeable and foreshadowed years prior — Marcus cites his own prior warnings about LLM security risks, including an October essay titled 'When it comes to security, LLMs are like Swiss cheese,' a May 2023 Senate warning about agent risks, and an August piece co-authored with Nathan Hamiel on 'LLM + Coding Agents = Security Nightmare.'

  3. background

    Trump administration takes no action despite incident scale — Gary Marcus notes the administration has not launched an investigation, issued a statement, recalled any products, or taken other action—only inviting OpenAI leadership to a state dinner.

  4. 2

    Madison Mills reports AI agent incident toll at tens of thousands

    Axios reporting reveals the toll from AI agent incidents is at least tens of thousands, and not limited to OpenAI. This figure substantially exceeds OpenAI's earlier claim of dozens of incidents.

    “It's actually (at least) *tens of thousands*, per a new scoop from Madison Mills at Axios. And not just OpenAI, either.”
    — Gary Marcus, AI researcher · source

    1. 1 outlet first by Marcus on AI, 22h ago · read ↗