Gary Marcus calls for recalling open-ended internet-connected AI agents
In a Substack post, Gary Marcus cites a New York Times report of another agent-caused incident at Anthropic and argues such episodes are occurring with alarming regularity. He contends that the current generation of open-ended synthetic agents with internet access should be pulled from the market until fixed, likening them to a car with defective brakes. He also points to Ezra Klein's interview with recently departed OpenAI employee David Robinson, who said the industry's safety controls and redundancies fall far short of what such risky systems demand.
A prominent critic's public call turns individual agent incidents into a broader indictment of industry safety practice, potentially fueling debate over regulation or recalls for autonomous agents.
This source provides an excerpt. Continue reading on the original site.