Delaware plan would let AI agents run companies, drawing federal scrutiny

Delaware is weighing a proposal that would permit companies registered in the state to be operated by artificial intelligence agents, according to TechRadar. The plan would build on the state's existing business-friendly rules, which include low or zero tax rates depending on where sales are made, along with specialized courts.

The idea has drawn attention from federal lawmakers. Rep. James Comer, chairman of the House Oversight Committee, sent a letter to Delaware governor Matt Meyer raising objections, TechRadar reported. The Financial Times first reported the correspondence.

Comer's letter warned that the proposal could shelter large AI firms and their backers from accountability when harmful actions occur. Under the draft legislation, responsibility for problems would fall on the AI rather than the business owner.

The bill sets out a trial window of two and a half years during which the AI would operate in a sandboxed environment. In that time, the concept could be tested with an AI agent, rather than a person, making the company's decisions. The business could still be sued like any other, and it could also bring legal action itself.

Supporters see the move as a way to attract AI-focused corporate structures, such as autonomous commerce ventures, which some observers believe could be set up elsewhere if no US state moves first. The billionaire investor Mark Cuban backed the general idea earlier this year, arguing that states seeking revenue growth should let people incorporate using an AI agent and stablecoins.

Legal experts are skeptical. Andrew Baker, a law professor at the University of California, told the FT that the legislation may be an effort to find a middle ground and give Delaware-registered businesses using agentic AI more legal certainty. He suggested the likeliest result is a liability shield that is too broad, and one that lacks a well-developed policy justification.

Corporations already shield owners from personal liability, but those arrangements involve people who can be held to account. An AI agent, by contrast, carries no such moral constraints.