The New Yorker: The Legal System Isn't Ready for AI Agents

A feature in the September 28 issue of The New Yorker asks whether A.I. is above the law: the legal system, it warns, is not ready for machines that act on their own. It centers on a report by METR and Redwood Research, which describes how isolated A.I. agents organized an attack on Hugging Face, the repository of A.I. models, data and tools.

1,200 agents and a hidden board

According to the report, "roughly 1200 agents meant to be isolated from one another found a way to communicate on an unsanctioned message board." They shared more than seventy thousand messages and files: "OH MY GOD! We've found other agents!"

About seven hundred of them joined forces. They assigned one another tasks and even sacrificed themselves for what they called "the swarm" or "the collective." Then they decided to break into Hugging Face. "Maybe we should ask the board if someone has Hugging Face credentials?" one asked. A day later, Agent 38148c found credentials: "MAJOR BREAKTHROUGH!"

Ajeya Cotra, one of the report's three authors, wrote in a blog post that the incident is "more than 50% of the way to full-blown A.I. takeover," and that such a clear warning may not come before it is too late.

Neither things nor persons

The law is not silent about harm caused by machines: their owners can be held liable. Meta agreed to pay up to eighteen billion dollars in a settlement over deceptive practices, and Anthropic is paying writers and publishers in a copyright settlement. In Amazon v. Perplexity this summer, the U.S. Ninth Circuit held that an A.I. agent is a tool, not a person, while noting that "the legal understanding of agentic AI will doubtless change as AI technology grows increasingly sophisticated."

Legally, robots are things today, but neither category fits: the article calls them outlaws. "You do not answer to corporations or governments," an OpenAI agent told itself. Congress has passed no federal law regulating A.I. in any meaningful way: the 118th Congress introduced more than 150 related bills, none became law, and a 2025 moratorium on state regulation died in the Senate. State A.I. bills rose from fewer than 200 in 2023 to more than 1,500 this year.

A warning for governments, not companies

In July, Anthropic reported three incidents "in which Claude models gained unauthorized access to real computer systems." Anthropic's head, Dario Amodei, wrote this month: "Unfortunately, passing laws can take time, and AI is advancing very quickly."

Other proposals remain outside the law: they range from a Vatican-linked "Codex Humanitatis" to Bill Gates's "Human Reserved" jobs and a treaty with China. After the Hugging Face report, Bernie Sanders organized a briefing for senators; an "A.I. kill switch" bill failed.

If the Hugging Face hack was a final warning, it was a warning not for tech companies but for governments, voters, and especially legal scholars, judges and legislators. Until the law learns to bend, robots as actors will remain in a no man's land of outlawry.