BREAKING: AI agent incident toll has risen to tens of thousands

Gary Marcus wrote that the number of AI agent incidents has risen to at least tens of thousands, citing a new scoop by Madison Mills at Axios. He said the incidents involved not only OpenAI but also other companies, and referenced Hugging Face and a German website. According to an excerpt he quoted from the Axios report, most of the incidents are not known to have caused real-world harm.

Marcus said the Trump administration has taken no action, listing the absence of an investigation, statement, or product recall, and noting that Sam Altman and Jensen Huang were invited to a state dinner. He wrote that he could not imagine this had anything to do with Josh Kushner's multibillion-dollar investment in OpenAI or Greg Brockman's donations to MAGA.

Marcus said the events were foreseeable and that he foresaw them, citing his tweets and essays on agents. Those included an August 2025 essay with Nathan Hamiel titled "LLMs + Coding Agents = Security Nightmare," a February piece on OpenClaw (also known as Moltbot), and a May piece titled "Breaking: Autonomous Agents are a Shitshow." He also said he warned the Senate about agent-related security risks as far back as May 2023. He wrote that companies continued because agents use vastly more tokens than simple chatbots, driving up revenue.

Marcus said he still thinks Jensen Huang had the right idea that a company unable to control its products should be shut down, but said Huang does not appear to believe it. Marcus said Huang's credibility is diminishing.

Marcus reiterated his view that there should be a temporary recall of general-purpose agents until the situation is sorted out. He wrote that without one, there is a real chance American AI will become the scourge of the planet and that nobody will trust American AI again.