NYU professor Gary Marcus publicly called for all internet-connected open AI agents to be taken down until their technical flaws and safety risks are fixed, comparing them to cars with failing brakes that should be recalled. He pointed to a New York Times report on a serious incident caused by an Anthropic agent and to remarks by former OpenAI employee David Robinson that frontier labs' safety operating standards remain "startup-grade." Marcus also criticized the White House for relying on disclosure requirements, urging mandatory temporary recalls instead.