US Congress questions Delaware proposal to allow AI agents to run companies
On October 26, US House Oversight Committee Chairman James Comer sent a letter to Delaware Governor Matt Meyer requesting documents related to the proposed "AI company" (AIC), including safety assessments and details of industry participation, and warned that the investigation could push federal legislation to restrict state-level AI initiatives. According to the draft, Delaware would establish a 30-month regulatory sandbox in which AICs could have AI agents handle business decisions without routine human oversight; the company itself could sue and be sued, and its members would generally be exempt from company liability, except in cases such as fraud or intentional violations of law. Andrew Baker, a law professor at the University of California, Berkeley, told the Financial Times that the liability exemption is too broad; Delaware Secretary of State Charuni Patibanda-Sanchez, together with Norm Ai CEO John Nay, presented the proposal in Fortune and mentioned that company formation fees could generate fiscal revenue. The bill is expected to be submitted next year.
Why read this
AI score 59
Delaware plans an AI company sandbox, putting liability attribution and immunity boundaries on the table; federal lawmakers have requested safety assessment materials.
Topics
Body
WTF?! The idea of an entire business being run by an AI agent sounds like a nightmare, but it's a scenario that Delaware is pushing for. Now, the state's plan is facing pushback from a senior lawmaker who says having an AI in charge could protect owners and investors from liability for harmful actions.
House Oversight Committee chairman James Comer wrote to Governor Matt Meyer on October 26, requesting documents about the proposed Artificial Intelligence Company, or AIC.
The Kentucky Republican wants details of safety assessments and industry involvement, warning that the investigation could lead to federal legislation restricting state AI initiatives.
Under the draft bill, Delaware would establish a 30-month regulatory sandbox to test the concept. An AIC could pursue any lawful purpose except banking, with an AI agent handling business decisions without routine human supervision. The company itself could sue and be sued.
Each AIC would have one member, either a person or an entity, responsible for keeping it adequately funded. That member would generally be protected from the company's liabilities. However, the draft includes exceptions for fraud, willful lawbreaking, and allowing activities the member knew, or should have known, fell outside the company's stated purpose.
The proposal would not make every AI-related disaster somebody else's problem. But critics question whether those exceptions offer enough protection when software makes the decisions and humans collect the profits.
"I'm afraid your weak, fleshy body is no longer required at this company"
UC Berkeley law professor Andrew Baker told the Financial Times that the proposed liability shield appeared excessively generous and lacked a sufficiently developed justification.
Supporters argue that giving autonomous systems a legal identity would make their actions traceable and provide a target for liability. Delaware Secretary of State Charuni Patibanda-Sanchez outlined the proposal with Norm Ai CEO John Nay in Fortune.
There's also a financial incentive, naturally. Spotlight Delaware reports that Patibanda-Sanchez highlighted the formation fees these companies could generate. Corporate taxes and fees already contribute more than $2 billion annually to Delaware's budget, roughly a third of its revenue.
The idea of putting AI in charge of a company isn't new. In October 2025, Logitech CEO Hanneke Faber called for AI agents in every board meeting. A month later, OpenAI boss Sam Altman said he wants his company to become the first big firm run by an AI.
For now, this remains a proposal. Spotlight Delaware reported that the bill was likely to be introduced next year, while Meyer's office told Ars it was reviewing Comer's letter.
Given the many recent incidents in which AI agents have gone rogue and hacked other companies, putting the technology in charge of a company seems like a rash move. That's without mentioning the repeated warnings from experts about an AI apocalypse. Or maybe it's exactly what the sneaky AI wants?