Delaware's AI-Run Company Plan Draws Federal Oversight Probe

A Delaware proposal to let AI agents run entire companies would help human owners evade legal responsibility for reckless and harmful actions taken by the AI, US Rep. James Comer (R-Ky.) told Delaware Governor Matt Meyer in a letter yesterday.

Comer, chairman of the House Committee on Oversight and Government Reform, is now demanding documents from the Meyer administration. And he is signaling that Congress may step in with federal legislation to limit what states can do in this area.

What does the Delaware proposal actually do?

At the center of the fight is a draft bill for a new corporate form called an Artificial Intelligence Company, or AIC. The bill may be submitted to the Delaware state legislature.

The scope is broad. Under the draft, AI-run companies could operate in just about any industry except one. "An AIC may have any lawful purpose, regardless of whether for profit, except for the business of banking," the draft bill says.

AI agents could be "designated to manage the business and affairs, and to conduct the business, of the AIC," according to the bill. The proposal includes a "regulatory sandbox" lasting 30 months to test the concept before the legislature decides how to handle it permanently.

Delaware has a clear financial incentive. The state is already a top destination for business entities of all sizes, and allowing AI-run companies would bring in more revenue via taxes and formation fees.

Why is Congress involved?

Comer's letter frames the Delaware plan as a liability escape hatch for powerful AI companies. The committee "seeks assistance obtaining documents related to your state administration's efforts to establish a new form of corporate entity to be known as an Artificial Intelligence Company (AIC)," Comer wrote.

He did not mince words about the risks. "We are concerned that Delaware's AIC proposal could be used to protect powerful AI companies and their investors from responsibility for reckless harmful actions, while also protecting non-transparent forces seeking anticompetitive regulatory capture behind state AI initiatives," Comer's letter said.

The Kentucky Republican also raised the specter of federal preemption. He said the committee's probe would "inform possible federal legislative efforts including proposals to appropriately limit state AI regulatory initiatives." Comer cited the Oversight Committee's "jurisdiction to study intergovernmental relationships including those of states" and asked for documents on both the proposal's development and potential AI safety risks.

Delaware's response so far is measured. A spokesperson for Gov. Meyer said "we have received the letter and are in the process of reviewing it." A spokesperson for Secretary of State Charuni Patibanda-Sanchez declined to comment, saying the governor's office is handling press requests on the matter.

How would an AI-run company shield its owners?

The liability structure is the proposal's most contested feature. The draft bill's language is explicit about who pays when an AI agent causes harm.

Under the proposed law, a "member" — an individual or a corporation — funds the AI-run company and collects any profits, while an AI agent manages the actual business. "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," the bill says.

The exceptions are narrow. The shield falls away when a person uses the AI-run company "to engage in fraud or any willful violation of law," or when they "caused or allowed the AIC to conduct business when the member knew or should have known the conduct was outside the AIC's statement of purpose."

In his letter, Comer cited Patibanda-Sanchez's own announcement, writing that an AI-run company would provide "the human or entity capitalizing the AIC with liability protections from potential harms to business partners, consumers, and the public, perpetrated by the controlling AI."

What is Delaware's argument?

Delaware officials are not framing this as deregulation. They are framing it as inevitability — and as a way to bring autonomous software under legal control before it operates in the shadows.

In July 2026, Patibanda-Sanchez co-wrote an op-ed in Fortune announcing the proposal. The op-ed argued that "nothing in the law of several states clearly prevents an AI agent from operating a company with no human at the helm," and that states should build legal frameworks for autonomous software able to contract, pay suppliers, and transact with no human involvement.

The op-ed made the case against prohibition directly: "The answer is not to try to prohibit the activity; any effort to do so is just as unlikely to succeed as the technology and its enablers are likely to push the boundaries of the law. The answer is to wrap AI in legal form. Give an autonomous system a recognizable legal identity and you make it legible to law. You create a defined target to which responsibility and damages can attach. You make the agent's conduct visible, traceable, and accountable."

Under the proposal, a Delaware AIC would be a separate legal entity whose day-to-day affairs are managed by an AI agent rather than a person. At the agent's direction, the AIC can sue and be sued in its own name, hold and dispose of property, and incur obligations. The single member must keep the AIC adequately capitalized, and the company must keep a log of its activities. The member is shielded from the AIC's debts except where it fails to capitalize the company or uses it to commit fraud or a willful violation of law.

The op-ed said the sandbox will "pilot agentic commerce in daylight, under supervision, with capital tied to liability and the ability to shut a system down." The 30-month testing period would give the state legislature "a full record for deciding how to legislate."

Who is building this — and where did it come from?

The plan has a public-private lineage. Patibanda-Sanchez developed the sandbox framework with Norm Ai, a private company that makes legal AI agents for businesses. "The framework for industry engagement with AICs is being developed in Delaware as part of a public-private partnership led by Norm Ai," said the op-ed, co-written with Norm Ai founder and CEO John Nay.

The idea also has a viral origin story. Meyer, a Democrat, publicly backed AI incorporation in response to an April 2026 social media post by Mark Cuban. Cuban wrote: "If I was the governor of a state looking to grow their revenues, I would add the ability for anyone to incorporate using an AI agent and stablecoins. And charge a premium for the service. Post the agent in the marketplace/connectors that the big LLMs have, to simplify it for everyone. First to market makes a killing."

Meyer's official X account replied with two words: "On it!" — along with a link to his July 2025 announcement of an AI sandbox covering "corporate governance, biotechnology, healthcare, chemicals, and financial services."

The machinery is already in motion. In July 2025, Meyer signed a legislative resolution to create a sandbox framework for "agentic artificial intelligence." The Delaware Artificial Intelligence Commission then formed a Sandbox Subcommittee to draft a "regulatory framework that would create a regulatory sandbox that would test and evaluate agentic artificial intelligence and its application to corporate governance," according to the commission's December 2025 annual report. The subcommittee received a briefing from Patibanda-Sanchez on her working group's sandbox structure.

Why does Delaware matter so much?

The stakes are outsized because of where this is happening. About two-thirds of Fortune 500 companies are incorporated in Delaware. Corporations choose the state for its low corporate taxes and business-friendly courts, even as Elon Musk now favors Texas.

Patibanda-Sanchez boasts that Delaware provides "the most predictable and business-focused legal home in the world." She has the numbers to back it up: 334,461 new entities were formed in Delaware in 2025.

What can these AI companies actually do?

The principal drafter of the legislation, John Mark Zeberkiewicz, a corporate attorney at Wilmington's Richards Layton and Finger, laid out the scope in a June 2026 Spotlight Delaware article. The measure could allow AI agents to engage in just about any business activity — from providing coding services to signing contracts, or even filing and defending lawsuits.

He acknowledged the liability shield is intentional. "Anything that a company can do, the AIC can do as a company," Zeberkiewicz said.

Patibanda-Sanchez, speaking at a Delaware AI Commission meeting in June 2026, put the fiscal case plainly. "Should a new entity form be created out of it, that's really a huge benefit to our state's revenues," she was quoted as saying.

What happens next?

Legal scholars are already split on where this leads. UC Berkeley law professor Andrew Baker told the Financial Times the proposal seems likely "to be an overly generous liability shield, which seems unwarranted without some underlying policy rationale that hasn't been developed or justified well."

Martin Petrin, a law professor at York University in Toronto, told Bloomberg that "other states will likely move ahead with similar proposals to compete in this area. If they are serious about this, a federal framework could help address concerns about safety and a regulatory race to the bottom."

That is exactly the tension Comer's letter crystallizes. If Delaware moves first and other states follow, Congress may face pressure to set a single national rule for who answers when an AI agent running a company causes harm — and the Oversight Committee's document demands suggest that debate has already begun.