House Oversight opens investigation into Delaware’s plan for AI-run companies

An AI agent could run a business while its human or corporate backer receives liability protections under a Delaware proposal now facing congressional scrutiny. House Oversight Committee Chairman James Comer opened an investigation on October 26, 2026, seeking records on the plan’s risks and the interests shaping it.

Comer asked Governor Matt Meyer to supply records by November 9, including risk assessments, outside consulting agreements, draft legislation and project communications dating back to January 2025. The request reaches internal AI chat logs, according to Insurance Business. It seeks both the state’s assessment of potential harm and information about the people and organizations influencing the project.

The investigation concerns a proposed Artificial Intelligence Company, or AIC—not a newly authorized business form. A 36-page draft was released in June 2026, and the bill is expected before Delaware lawmakers in their next legislative session, Insurance Business reported.

The company, rather than its backer, bears the debt

Under the proposed structure, an AI agent would manage the company’s daily affairs. The AIC would be a separate legal entity able to hold property, enter obligations, and sue or be sued in its own name. The design would give autonomous software a way to conduct business through a legally recognized company, rather than treating the software itself as the contracting party.

The AIC would have one member—a person or another entity—responsible for keeping it adequately funded. That member would generally be protected from the company’s debts. Exceptions would apply if the member failed to provide adequate capital or used the company for fraud or a willful violation of law. The proposed protection is therefore conditional, not an unrestricted exemption from responsibility.

That allocation has consequences for anyone seeking compensation after a loss. Insurance Business’s analysis says a claimant might be limited to recovering from the AIC’s own assets and insurance when its backer is protected. Capital requirements and activity records would consequently become important to assessing whether an injured party could recover money and reconstruct what the agent did.

A supervised trial, with powers to shut companies down

The proposal’s advocates describe a controlled experiment rather than permission for unrestricted autonomous commerce. In a July 14 article published by development partner Norm Ai, its CEO John Nay and Delaware Secretary of State Charuni Patibanda-Sanchez set out a regulatory sandbox: a supervised, temporary testing program.

• Admission would be decided by a committee including state officials, outside attorneys and technologists. Participating companies would have to meet capital requirements and keep activity logs.
• Business counterparties would receive notice that the AIC is a temporary test entity, lacks state endorsement, and has a specified end date and complaint process.
• Officials could suspend an AIC, revoke its authorization or seek its dissolution in court. Banking would be excluded, and the program would expire after 30 months.

Nay and Patibanda-Sanchez argued that giving autonomous commerce a recognizable legal home would make its conduct traceable and provide a defined target for damages. They said consumer-protection and criminal law would continue to apply in full. Their case is that a supervised domestic trial would be more accountable than allowing the activity to migrate offshore or onto anonymous infrastructure.

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

House Oversight Committee Chairman James Comer, in his letter to Governor Matt Meyer

Comer wants the policymaking record

The inquiry challenges both the liability arrangement and the transparency of the process that produced it. Comer pointed to the Delaware AI Commission’s December 22, 2025 report, which described work on a sandbox for AI in corporate governance. He said that report did not mention assessments of potential safety problems, harms or risks, or consultation with independent experts. That criticism concerns the report’s omissions; it is not a finding that no assessments occurred.

The committee also questioned the visibility of industry participation. Its release cited September reporting that the Secretary of State’s office had declined to answer questions about its working relationship with an AI company contributing to the proposal. Norm Ai’s public partnership announcement explains its stated role in developing the framework, but Comer is seeking the underlying agreements and communications.

The inquiry could reach beyond Delaware’s draft. Comer said the requested information could inform possible federal legislation limiting state AI regulatory initiatives. For now, the concrete federal action is a demand for records—not a prohibition on Delaware’s proposal or a decision on whether its safeguards are sufficient.