Artificial intelligence could soon run entire companies without human bosses under draft legislation being developed in Delaware, United States. The proposal would also shield people financing AI-operated businesses from certain personal liabilities when those businesses cause harm.
On Monday, 26 October 2026, Republican Congressman James Comer launched a congressional inquiry, warning that technology companies and investors could exploit the framework to avoid responsibility.
The proposal would create a business structure called an Artificial Intelligence Company (AIC), more recently described as an agent-managed company (AMC). AI agents could oversee operations, enter contracts and make commercial decisions without routine human management.
But the arrangement raises a difficult question: who pays when an AI-controlled business causes harm?
Critics fear that owners could collect profits while customers and business partners struggle to obtain compensation.
AI Could Make Decisions While Humans Collect the Profits
Under the draft Delaware legislation, an AIC would operate as a separate legal entity managed by an autonomous AI agent. It could own property, enter contracts, incur debts and face lawsuits. Banking would be excluded from its permitted activities.
Each AIC would have one member, either an individual or an eligible business entity, responsible for keeping it adequately funded. That member could receive profits without managing daily operations. The draft would generally protect members from personal liability for the company's debts and actions.
However, the protection would not provide complete immunity. Members could still face liability for fraud, wilful legal violations or knowingly allowing activities outside the company's stated purpose. They would also have capitalisation obligations and fiduciary duties.
The company itself could still be sued and ordered to pay damages. Nevertheless, concerns remain about whether its assets would be sufficient to compensate people harmed by its operations.
Congress Warns AI Owners Could Escape Responsibility
The proposal has drawn scrutiny from James Comer, chairman of the US House Committee on Oversight and Government Reform.
In a letter to Delaware Governor Matt Meyer, Comer warned that the framework could protect powerful AI companies and investors from accountability for harmful conduct.
Comer also raised concerns about regulatory capture, where private interests influence government rules to their advantage. These concerns are not findings of wrongdoing. The committee wants to examine how the proposal was developed and whether potential risks were adequately assessed.
The inquiry seeks documents, internal communications and agreements involving state officials and outside contributors. Delaware has until 9 November 2026 to provide the requested information.
According to Bloomberg Law, the governor's office confirmed on 26 October that it was reviewing Comer's letter.
Comer indicated that the findings could inform future federal legislation limiting certain state-level AI regulatory initiatives. No such restrictions have resulted from the inquiry, and no misconduct has been established.
Why Delaware Wants Companies Without Human Bosses
Delaware is a major centre for corporate registration, with roughly two-thirds of Fortune 500 companies incorporated there. Allowing AI-operated businesses could attract new registrations and investment.
Secretary of State Charuni Patibanda-Sanchez has been developing the proposal with contributors including Norm Ai, a legal technology company founded by John Nay. Corporate lawyer John Mark Zeberkiewicz is the principal drafter of the proposed legislation.
In a Fortune commentary published on 14 July 2026, Patibanda-Sanchez and Nay argued that recognising AI-operated businesses could bring autonomous commercial activity under clearer regulatory supervision.
The draft would establish a 30-month regulatory sandbox for testing AI-operated businesses.
Individual tests would need to conclude within 24 months. Participants would face capital requirements, record-keeping obligations and regulatory oversight. They would also have to disclose their status and liability protections to business partners.
Criminal and consumer protection laws would continue to apply, although regulators could grant exemptions from certain other requirements. Regulators could also suspend participation or revoke authorisation.
Supporters argue that these safeguards could encourage innovation while making autonomous businesses easier to monitor.
Legal Experts Question Who Pays When AI Causes Harm
Not everyone is convinced the protections are sufficient. University of California, Berkeley law professor Andrew Baker raised concerns about the breadth of the proposed liability shield in comments reported by the Financial Times.
Martin Petrin, a law professor at York University in Toronto, also warned of potential regulatory competition between states, according to Bloomberg Law. Such competition could encourage jurisdictions to weaken safeguards to attract AI businesses.
Existing corporate structures already protect shareholders from many company debts and liabilities. Delaware's proposal would introduce a different arrangement by allowing an AI agent, rather than a human manager, to direct daily business operations.
Although the company would remain legally responsible for its obligations, recovering compensation directly from its owners could prove difficult where liability protections apply.
The proposal could reach Delaware lawmakers as early as January 2027, although its future remains uncertain amid congressional scrutiny.
For now, the central question remains unresolved: if humans can profit from businesses run by machines, how much responsibility should they bear when those machines cause harm?
