On October 26, 2026, the U.S. House Committee on Oversight and Government Reform announced that Chairman James Comer had asked Delaware Gov. Matt Meyer for documents and information about risks assessed for the state’s proposed Artificial Intelligence Company (AIC) framework and the forces driving similar initiatives. The committee said the material could inform possible federal legislation.
On October 27, a spokesperson for Meyer said the governor’s office had received the letter and was reviewing it.
What the House committee wants from Delaware
The committee’s request focuses on how Delaware assessed risks tied to the proposal and what is driving Delaware and other states to pursue similar frameworks. Comer said the committee was concerned the proposal could shield powerful AI companies and their investors from responsibility for harmful actions, while opaque forces could use state initiatives to pursue anticompetitive influence.
What an AI-managed company would be
A June 2, 2026, draft bill defines the AIC as the proposed company and an AI agent as the software manager. The agent would make decisions and act toward predetermined goals without routine human supervision or intervention. The company, not the software, would have one member, who could be an individual or a specified Delaware entity. That member would provide capital and identify a member representative.
The draft would allow an AIC to pursue any lawful purpose except banking. If an AIC had no manager, its member would manage it in the meantime and have 30 days to designate a replacement. The proposed structure therefore places an autonomous agent in the management role while retaining a member with defined responsibilities.
The draft’s safeguards and liability limits
Each proposed innovation would require a separate application and could be tested only with a counterparty. Applicants would describe expected public benefits, risks to counterparties and others, data-security measures, testing and wind-down plans, and capitalization. They would also have to show they could terminate the AI agent. Describing a public benefit would be an application requirement, not a guaranteed result for residents.
The draft would limit the sandbox committee to approving regulatory exemptions an applicant specifically requested. If a relevant state agency refused an exemption under its authority, that refusal would bind the committee. The committee could suspend or revoke participation and order an AIC dissolved on specified grounds.
Before a counterparty interacted with an authorized AIC, the company would disclose the member’s identity and contact information, the test’s temporary nature, related liability protections, Delaware’s non-endorsement, and complaint channels. The draft also proposes that covered disputes be heard in Delaware’s Court of Chancery.
The proposed liability shield would generally protect a member from AIC debts solely by virtue of membership. It would also limit liability for certain acts taken by the AIC, or at its direction, solely under agency or instrumentality theories when the AIC was adequately capitalized. The draft preserves exceptions for fraud, willful violations of law, and allowing activity beyond the AIC’s stated purpose when the member knew or should have known it was outside that purpose. Sandbox participants would not receive immunity from civil damages and would remain subject to Delaware criminal and consumer-protection laws.
Three different time limits
The June 2 draft sets three proposed periods with separate triggers: the sandbox program would run for 30 months after implementation; each participant would have 24 months after receiving a registration number to complete its test; and the act would expire 36 months after enactment unless the General Assembly extended it.
