The Pulse
Comer Seeks Records on Delaware’s AI-Run Company Plan
House Oversight Chairman James Comer has asked Delaware Gov. Matt Meyer for records on a proposal to let AI agents manage companies. The request raises questions about liability protections, safety reviews and the role of private-sector par
AI.info Team · Written with AI · fact-checked ·
“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.”
James Comer, chairman, House Committee on Oversight and Government Reform
Rep. James Comer sent the warning to Delaware Gov. Matt Meyer on Monday, October 26, as the House Oversight and Government Reform Committee opened an inquiry into the state’s plan for artificial intelligence companies. The proposed entities, called AICs, would put AI agents in charge of day-to-day business operations while a human or organization supplies capital. Comer says the committee wants records on how Delaware assessed the proposal’s risks and who helped shape it.
Comer sets a November 9 deadline
The four-page letter asks Meyer’s administration for documents and communications dating to January 1, 2025, including internal AI chat logs. It also seeks outside contributors’ ethics and consulting agreements, records of safety-risk discussions, and drafts and comments on AIC policy proposals. The committee requests the material by November 9, 2026.
Comer says the inquiry may inform possible federal legislation to limit state AI initiatives. The letter points to a December 2025 report from Delaware’s Artificial Intelligence Commission that described work on a regulatory sandbox for testing AI in corporate governance, but did not mention safety or harm assessments or consultation with independent experts. Those omissions are the committee’s stated concern; the letter does not establish that Delaware failed to conduct additional reviews.
The draft would give an AI agent control of company affairs
A June 2 draft bill would let an AI agent manage an AIC’s business and affairs, signifying a corporate structure in which a software agent—not a human director—makes routine decisions. The entity could have a lawful purpose other than banking and could hold property, incur obligations, and sue or be sued. The draft is a proposal, not an enacted law.
Under the draft, the AIC’s sole member would fund the company and keep it adequately capitalized. The member would generally not be personally liable for the AIC’s debts or for actions taken at the AI-run company’s direction if the company was adequately capitalized. The draft preserves liability in specified circumstances, including fraud, a willful violation of law, or knowingly allowing the company to act outside its stated purpose.
The bill would also create a 30-month sandbox, with a state committee overseeing admission and tests. The draft describes application requirements that include assessments of risks to the public and counterparties, a plan for winding down a test, and proof of adequate capitalization. The scope of the proposed liability protections, and whether those safeguards would be enough in practice, sit at the center of the congressional inquiry.
Delaware’s sponsors describe a supervised test
Delaware Secretary of State Charuni Patibanda-Sanchez and Norm Ai founder John Nay presented the proposal in a July Fortune commentary as a way to bring autonomous business activity under legal rules. They argue that a separate legal entity, activity logs, capital requirements, and a supervised sandbox could make an agent’s conduct traceable and provide a target for claims. Their account also says officials could suspend an AIC or revoke its authorization.
The partnership itself is part of the oversight question: the commentary says Norm Ai leads the public-private effort to develop industry engagement with AICs. Comer’s letter asks for records on outside contributors and on how the state considered safety and liability; the inquiry is not a finding of wrongdoing. Meyer’s office told Ars Technica it had received the letter and was reviewing it.
The unresolved issue is who pays when an agent causes harm
The proposal separates the AI agent’s operational authority from the member’s personal liability, subject to the draft’s conditions and exceptions. That structure may offer a legal route for testing agent-run businesses, while also leaving consumers, counterparties and regulators to determine what remedies apply when an agent’s decisions cause losses. The committee’s request now puts the proposal’s drafting history, safety review and liability design before Congress as Delaware considers a plan it has not yet enacted.