Delaware is developing a legal structure that would allow an AI agent to run a company with no routine human supervision, and now a sitting House committee chair wants to know exactly who authorized that idea. On October 26, 2026, House Oversight Committee Chairman James Comer sent Delaware Gov. Matt Meyer a formal document request covering risk assessments, draft legislation, outside participants, safety reviews, and reportedly internal AI chat logs related to the proposal.
The central question Comer raised is not theoretical. If an autonomous system makes a decision that harms a consumer or triggers a regulatory violation, who actually pays?
What Delaware Is Actually Proposing
Delaware’s draft would create an entirely new corporate category, the Artificial Intelligence Company, built for autonomous operation from the ground up.
Under the proposal, an AI agent would direct a company’s day-to-day business with no requirement for routine human decision-making. The entity would have one member, either a person or an organization, responsible for keeping it adequately funded.
That member would hold broad liability protection from actions taken by the company or by anyone operating at the AIC’s direction. The AIC itself could own property, enter contracts, and sue or be sued, and it could pursue any lawful purpose except banking.
The framework would reportedly run inside a 30-month regulatory sandbox overseen by state officials and outside experts, according to Bloomberg Law reporting. AICs would reportedly also have to disclose their AI-operated status to counterparties, though that requirement has not been confirmed in primary draft language.
None of this is law yet. Delaware’s Artificial Intelligence Commission developed the framework following a resolution Gov. Meyer signed in July 2025 to test agentic AI in corporate governance. Legislation could be introduced as early as 2027, and Meyer’s office said it is reviewing Comer’s letter.
The Accountability Gap Congress Can’t Ignore
A liability shield built for human executives runs into a conceptual wall when the decision-maker is software.
The draft reportedly protects the AIC’s member from acts or omissions by the company unless fraud, undercapitalization, willful unlawful conduct, or negligent operation outside the company’s stated purpose is involved. How courts would apply standards like “willful” or “knowing” to an AI system’s output remains an open and unresolved policy question.
Proponents of the AIC structure argue it addresses a practical problem. Giving an autonomous entity a distinct legal identity means regulators and plaintiffs have a defined target, rather than chasing diffuse responsibility across a development team, a model vendor, and an investment vehicle.
The countervailing concern, which Comer raised directly, is that investors could collect the upside while deflecting responsibility to an algorithm when something breaks. Comer’s committee statement said the proposal could protect AI companies and investors from responsibility for “reckless harmful actions.”
His document request covered not just policy drafts but reportedly internal AI chat logs from the state’s own development process. That scope suggests the inquiry is examining governance practices as much as legal theory.
What Comes Next Matters Beyond Delaware
A state-level experiment in AI governance has become a federal policy question.
Comer noted the inquiry could inform federal legislation affecting state AI initiatives, which means the stakes extend well past Dover. Congressional pressure could reshape disclosure requirements, prompt calls for independent safety reviews, or open a broader debate over whether individual states should establish legal frameworks for autonomous AI businesses at all.
For anyone who might one day sign a contract with, buy from, or work for an AI-operated company, the unresolved issue is not who runs the business. It is who answers for it when something goes wrong.
