Delaware AI Company Plan Puts Agent-Run Firms Under House Scrutiny

Delaware built its reputation as the place where American corporate law gets written. Now the state is testing the Delaware AI Company proposal, an idea that would hand the daily running of a company to an artificial intelligence agent, and Washington has noticed. On October 26, House Oversight Committee Chairman James Comer sent a four-page letter to Delaware Governor Matt Meyer demanding documents about the proposal, with a deadline of November 9. The clash sets up one of the sharpest questions yet for the agent economy: when an AI runs the business, who answers for what it does?

The Delaware AI Company proposal would create a new corporate form called an Artificial Intelligence Company, or AIC. According to the House Oversight Committee, an AI agent would run the day-to-day operations of such an entity while the human or company behind it would receive liability protections from harms caused by the controlling AI. The committee said the information it gathers could inform possible federal legislation to limit state AI regulatory initiatives.

What a Delaware AI Company Would Be

Under the Delaware AI Company draft legislation, an AIC would be a separate legal entity whose daily affairs are managed by an AI agent rather than a person. At the agent's direction, the company could hold property, sign contracts, take on debt, and sue or be sued in its own name. The entity could pursue any lawful purpose except banking, according to reporting on the draft bill, and it could operate for profit or not.

The liability structure is the heart of the proposal. The individual or company that funds an AIC and receives its profits would, in principle, not be responsible for the corporation's acts and debts. Exceptions would apply where the funder failed to capitalise the company adequately or used it to commit fraud or a wilful violation of law. Supporters frame the design as agentic commerce conducted in daylight: the AI's activities would be logged, capital would be tied to liability, and the system could be shut down.

Delaware plans to test the Delaware AI Company concept for 30 months inside a regulatory sandbox before it becomes permanent law. The sandbox would let the state observe how agent-managed companies behave in practice, including how their records hold up when something goes wrong. The draft is expected to be introduced in the Delaware General Assembly in its January session, and it has no bill number yet.

Why Congress is asking questions

Comer's letter asks Delaware to hand over risk assessments, outside consulting agreements, draft bills, and communications on the project going back to January 2025. The request extends to internal AI chat logs, ethics agreements, and records of safety concerns. According to Digital Today, the committee is also examining who shaped the proposal and whether outside contributors were properly disclosed.

The chairman's central concern is that the Delaware AI Company liability shield could become a way for powerful AI companies and their investors to avoid responsibility for reckless or harmful actions taken by an AI. In the letter, Comer warned that the proposal could also shield non-transparent forces seeking anticompetitive regulatory capture behind state AI initiatives. The committee further flagged the role of John Mark Zeberkiewicz, a lawyer at Richards, Layton and Finger who has represented Facebook and has been described as the bill's principal drafter.

Delaware Secretary of State Charuni Patibanda-Sanchez has defended the direction of the project. In a July opinion piece, she argued that autonomous AI should be managed within a legal framework rather than banned, and the project has been developed in cooperation with legal AI developer Norm AI. Delaware AI Company supporters add that states should act before offshore alternatives set the terms instead. Critics, including law professor Andrew Baker of the University of California, see a possible good-faith push for legal certainty paired with what Baker described as an overly generous liability shield.

What happens next for agent-run firms

The stakes extend well beyond one small state. Delaware incorporates more than two-thirds of Fortune 500 companies, so a new entity form created there would ripple through corporate America. Insurers are already studying what the proposal would mean for directors and officers, cyber, and errors and omissions coverage, since a statute would settle part of the question of who pays when an autonomous system causes a loss, largely in favour of the people behind the company.

The Delaware AI Company debate also arrives as the infrastructure for trusted agents is maturing. Recent work on agent standards, sandboxed execution, and scoring the trustworthiness of agents shows the industry wrestling with the same accountability gap Delaware is trying to legislate. Readers can follow related coverage in the Agentic AI Foundation's agent standards, Akamai's agent sandbox for enterprise agents, and Socure's agent trust tools for businesses.

For now, the proposal remains just that: a draft, not law. Delaware must answer Congress by November 9, and the General Assembly will decide next year whether agent-managed companies become a real corporate form. Whatever it decides, the questions Comer raised will follow every agent that signs a contract, moves money, or makes a decision in a company's name: who is accountable, what records prove it, and how much protection the humans behind the machine should get. According to Bloomberg Law, the congressional scrutiny comes at a turning point for technology that is rapidly transforming business life, and the Delaware AI Company outcome may set the template other states follow.

Sources: House Oversight Committee press release, Bloomberg Law, Digital Today, and Ars Technica reporting on the draft bill.