Delaware draft bill would let AI agents manage companies, with owners shielded from liability

Delaware is preparing a new kind of legal entity, the Artificial Intelligence Company (AIC), in which an AI agent manages the business in place of a person. A draft bill says the human or corporate owner would not be liable for what the AIC does. On October 26, US Rep. James Comer (R-Ky.), chairman of the House Committee on Oversight and Government Reform, wrote to Delaware Governor Matt Meyer asking for documents on the plan, Ars Technica reports.

The question matters beyond one state. About two-thirds of Fortune 500 companies are incorporated in Delaware, so the rules it writes for corporate form tend to spread. This proposal sets out who answers for a business when software makes the decisions.

What the draft bill says

According to the draft bill quoted by Ars Technica, AI agents could be "designated to manage the business and affairs, and to conduct the business, of the AIC." An AIC "may have any lawful purpose, regardless of whether for profit, except for the business of banking."

Each AIC would have a single member, an individual or a corporation, that funds it and takes the profits. The draft says: "The member is not liable for any act or omission that the AIC takes or fails to take, or that another person takes or fails to take at the direction of the AIC." There are exceptions. The shield does not apply when the member uses the AIC "to engage in fraud or any willful violation of law," or let it conduct business when the member "knew or should have known the conduct was outside the AIC's statement of purpose."

The principal drafter, corporate attorney John Mark Zeberkiewicz of Richards Layton and Finger, told Spotlight Delaware in June that AI agents could do almost any business activity, from providing coding services to signing contracts and filing or defending lawsuits. "Anything that a company can do, the AIC can do as a company," he said. Spotlight Delaware reported that the bill is likely to go to the state legislature next year.

The AIC would first exist only inside a regulatory sandbox that lasts 30 months. During that period, the state would test the idea and collect a record for legislators.

The case Delaware makes

The plan was set out in July in a Fortune commentary co-written by Delaware Secretary of State Charuni Patibanda-Sanchez and John Nay, founder and CEO of Norm Ai, a company that makes legal AI agents for businesses. The authors say Norm Ai leads the public-private partnership that is developing the framework for industry engagement with AICs.

Their argument is that nothing in several states' laws clearly prevents an AI agent from running a company with no human at the helm, so prohibition would fail. "The answer is to wrap AI in legal form," they wrote. A recognisable legal identity would create "a defined target to which responsibility and damages can attach."

In their design, an AIC can sue and be sued in its own name, hold property and incur obligations. It must keep a log of its activities. The member must keep it adequately capitalised. If the member fails to do so, or uses the AIC for fraud or a willful violation of law, the member loses the shield against the company's debts. Officials could suspend an AIC, revoke its authorisation and ask the Court of Chancery to dissolve it.

Revenue is part of the motive. At a June meeting of the Delaware AI Commission, Patibanda-Sanchez said a new entity form would be "a huge benefit to our state's revenues," according to Spotlight Delaware. Meyer, a Democrat, signed a resolution in July 2025 to create a sandbox for agentic AI. In April 2026, Mark Cuban posted that a state should let anyone incorporate using an AI agent and stablecoins. Meyer's official X account replied, "On it!"

The objections

Comer's letter says the proposal "could be used to protect powerful AI companies and their investors from responsibility for reckless harmful actions." He says the committee's inquiry could "inform possible federal legislative efforts including proposals to appropriately limit state AI regulatory initiatives." He asked for documents on how the proposal was developed and on possible AI safety risks.

Legal scholars have raised concerns too. UC Berkeley law professor Andrew Baker told the Financial Times the plan seems likely "to be an overly generous liability shield" that lacks a well-developed policy rationale. Martin Petrin, a law professor at York University in Toronto, told Bloomberg that other states will probably propose similar entities to compete. He said a federal framework could help address safety concerns and "a regulatory race to the bottom."

A spokesperson for Meyer told Ars Technica that the office had received the letter and was reviewing it. Patibanda-Sanchez's office declined to comment.

What it means for work

The AIC would make it legal for software to do the work of a company's management: entering contracts, paying suppliers, making transactions and handling litigation, all without a human approving each step. The member's role shrinks to funding the company and defining its purpose, so the main human duties become capitalisation and setting scope, not daily decisions. Customers, suppliers and courts would deal with an agent as a counterparty, and the AIC's activity log would replace the memory and judgement of named managers as the record of what happened. Who carries the cost when an agent causes harm through error rather than fraud is still unresolved, and the sandbox is meant to show whether the capital and logging rules are enough to answer that. Whether the bill passes, how other states respond and whether Congress steps in will decide whether this becomes a common way to set up a business.