"No human at the helm": Delaware plan would let AI agents run entire businesses

Delaware’s bold push to let artificial‑intelligence agents run entire businesses has ignited a firestorm in Washington, with Republican Representative James Comer (R‑KY) warning that the proposal could become a legal shield for reckless AI‑driven enterprises. In a letter to Governor Matt Meyer, Comer, who chairs the House Committee on Oversight and Government Reform, demanded documents on the state’s effort to create a new corporate form dubbed the Artificial Intelligence Company, or AIC. His concerns echo a broader, bipartisan unease about granting autonomous software the power to make commercial decisions while insulating human owners from liability.

The Legislative Blueprint: What an AIC Would Look Like

The draft bill circulating in Delaware’s legislature outlines a corporate entity that can pursue any lawful purpose—except banking—while delegating day‑to‑day management to an AI agent. The language specifies that the AI “may be designated to manage the business and affairs, and to conduct the business, of the AIC.” A 30‑month “regulatory sandbox” would allow the state to observe AI‑run firms in real‑world conditions before deciding on permanent legislation. During this period, the AIC would be required to keep a log of its activities, providing a paper trail for oversight.

Crucially, the bill frames the human or corporate “member” that funds the AIC as a separate legal party. The member must keep the AIC adequately capitalized, but the draft explicitly states the member “is not liable for any act or omission that the AIC takes or fails to take… at the direction of the AIC,” except where fraud, willful law‑breaking, or reckless knowledge of the AI’s conduct is involved. This carve‑out is the very hinge on which Comer’s objections turn.

Delaware’s Rationale: Revenue, Innovation, and Legal Certainty

Delaware officials tout the proposal as a way to capture new tax and fee revenue while cementing the state’s reputation as a business‑friendly haven. The op‑ed co‑written by Secretary of State Charuni Patibanda‑Sanchez and Norm Ai founder John Nay argues that “nothing in the law of several states clearly prevents an AI agent from operating a company with no human at the helm.” By giving autonomous systems a recognizable legal identity, the authors claim the state can make AI conduct “visible, traceable, and accountable,” thereby turning a legal gray area into a regulated market.

Patibanda‑Sanchez points to Delaware’s existing corporate magnetism—about two‑thirds of Fortune 500 firms are incorporated there and more than 330,000 new entities were formed in 2025—as evidence that the sandbox could attract a flood of AI‑driven startups. The promise is that, under supervision, these firms could operate “in daylight” while the state gathers a “full record for deciding how to legislate” AI‑run companies.

Federal Pushback: Comer’s Letter and Potential Legislation

Comer’s letter to Governor Meyer frames the AIC as a “ripe for abuse” scheme that could let powerful AI firms and their investors dodge responsibility for harmful outcomes. He warns that the proposal could “protect non‑transparent forces seeking anticompetitive regulatory capture behind state AI initiatives.” By invoking the Oversight Committee’s jurisdiction over intergovernmental relationships, Comer signals an intent to draft federal legislation that would limit states’ ability to create such entities.

The representative’s request for documents on Delaware’s AI sandbox and safety assessments underscores a broader congressional concern: without federal standards, a patchwork of state‑level AI corporate forms could create regulatory arbitrage, allowing companies to cherry‑pick the most permissive jurisdiction. Comer explicitly ties the Delaware probe to “possible federal legislative efforts including proposals to appropriately limit state AI regulatory initiatives.”

Liability Shield or Legal Loophole? The Core Controversy

At the heart of the debate is whether the AIC structure merely extends Delaware’s well‑known corporate shield or creates a new loophole that could insulate owners from accountability for AI‑driven harms. The draft bill’s language that the member “is not liable for any act or omission… at the direction of the AIC” mirrors the limited liability protections already afforded to shareholders of traditional corporations. However, the addition of an autonomous decision‑maker raises the stakes: an AI could make rapid, high‑impact choices without human oversight, potentially causing consumer harm, market disruption, or environmental damage.

Comer’s concern is that the liability shield could be exploited by “powerful AI companies and their investors” to avoid responsibility for reckless or harmful actions. The bill does carve out exceptions for fraud, willful violations, or when a member “knew or should have known” the AI’s conduct was outside the stated purpose. Yet the subjective “should have known” standard may be difficult to enforce, especially when AI systems evolve beyond their original programming.

Industry Reaction and the Role of Private Partnerships

The proposal is not emerging in a vacuum. Patibanda‑Sanchez’s op‑ed notes a public‑private partnership with Norm Ai, a company that builds legal AI agents for businesses. By collaborating with a private firm, the state is effectively leveraging industry expertise to shape the regulatory framework. This partnership could accelerate the development of compliant AI agents, but it also raises questions about regulatory capture—whether the very firms that stand to benefit from the sandbox are influencing its rules.

Norm Ai’s involvement underscores a growing market for AI‑driven legal infrastructure. If the sandbox succeeds, it could spawn a new ecosystem of AI‑managed enterprises, from supply‑chain coordinators to autonomous service providers. Yet the same ecosystem could also produce “non‑transparent forces” that Comer warns about, especially if AI agents are programmed to prioritize profit over public safety.

Political Context: Delaware’s Leadership and the Push for Innovation

Governor Matt Meyer, a Democrat, has publicly supported the sandbox, echoing a social‑media post by billionaire Mark Cuban that suggested states could boost revenue by allowing AI‑incorporated firms and charging premiums for the service. Meyer’s X account responded with “On it!” and linked to a July 2025 announcement of the AI sandbox plan. This political backing reflects Delaware’s broader strategy to stay ahead of the regulatory curve and attract cutting‑edge businesses.

However, the governor’s office has so far offered only a brief acknowledgment of Comer’s letter, stating that it is “in the process of reviewing it.” The lack of a detailed response leaves open the question of how Delaware will balance the lure of new revenue against the potential for legal and ethical fallout.

What Lies Ahead: The Sandbox Timeline and Possible Outcomes

The 30‑month sandbox, slated to begin once the draft bill clears the legislature, will serve as a live testbed for AI‑run companies. During this period, participating AICs must maintain activity logs, remain adequately capitalized, and operate under the supervision of state officials. The sandbox’s “daylight” approach aims to generate a comprehensive record that can inform future legislation, whether to expand, modify, or abandon the concept.

Should the sandbox demonstrate that AI agents can manage businesses responsibly, Delaware could become the first state to grant legal personhood to autonomous software, potentially reshaping corporate law nationwide. Conversely, if the trial reveals significant harms—consumer losses, regulatory evasion, or anticompetitive behavior—Congressional action, spurred by Comer’s oversight committee, could curtail the model and impose federal limits on state‑level AI corporate entities. Either way, the outcome will set a precedent for how America reconciles rapid AI advancement with the rule of law.

This article was produced with AI-assisted research and editorial support. Reporting is based on the source material cited below. Sources: Ars Technica; arstechnica.com; Global1.News (28 October 2026).

By Jessica Ali, Staff Writer