Plan for AI ‘agents’ to run US companies challenged by lawmakers
Delaware is drafting a law for 'Artificial Intelligence Companies' (AICs): companies run day to day by autonomous AI agents that can sign contracts, sue and make decisions, inside a 30-month 'regulatory sandbox' pilot. House Oversight chair James Comer (Republican) wrote to Delaware governor Matt Meyer (Democrat) on Monday asking for documents by 9 November, warning the plan 'could be used to protect powerful AI companies and their investors from responsibility for reckless harmful actions'. He named John Mark Zeberkiewicz, a corporate lawyer who has represented Facebook, as the apparent 'principal drafter'. The proposal was developed by secretary of state Charuni Patibanda-Sanchez with Norm AI; in a July op-ed she framed it as running 'agentic commerce in daylight, under supervision, with capital tied to liability and the ability to shut a system down'. The draft shields individuals from personal liability if an AIC is adequately capitalised and its activities are logged; AI agents would be barred from banking, and no other business types have been chosen. Mark Cuban backs the idea ('First to market makes a killing'). Law professor Andrew Baker (UC) sees a possible good-faith push for 'legal certainty' but an 'overly generous liability shield'. Backdrop: Delaware is defending its incorporation franchise after Tesla, Dell and Dropbox moved to Nevada or Texas. Caveats: this is a proposal only, likely to be introduced in the January session, with no bill number yet; the FT names only Comer as signatory; and supporters argue states should act before offshore alternatives do.
Why it matters
If Delaware, which is fighting to keep its incorporation lead, creates a legal wrapper for agent-run companies, it would set a template for how AI-native start-ups are incorporated, who is liable when agents act, and what capital they must hold. A congressional probe this early shows the liability shield will be the political fight. For founders and investors in agentic AI, it is a signal to watch through the January session, not a regime that exists yet.
Executive takeaways
Proposal: Delaware 'Artificial Intelligence Companies' (AICs) run day to day by AI agents that can contract, sue and decide, in a 30-month regulatory sandbox. | • Challenge: House Oversight chair James Comer (R) asked Gov. Matt Meyer (D) for documents by 9 November, warning of shielding AI firms and investors from liability. | • Drafter: Comer flags corporate lawyer John Mark Zeberkiewicz (has represented Facebook) as apparent principal drafter. | • Sponsors: secretary of state Charuni Patibanda-Sanchez built it with Norm AI; pitch is supervised pilots with 'capital tied to liability' and a kill switch. | • Liability: individuals shielded if the AIC is adequately capitalised and activities logged; agents barred from banking. | • Voices: Mark Cuban backs it; Prof Andrew Baker calls the shield 'overly generous'. | • Timing: likely introduced in the January legislative session; no bill number yet. | • Context: Delaware is defending its incorporation lead after Tesla, Dell and Dropbox left for Nevada or Texas.
What Financial Times may be missing
The FT gives no bill text or definitions: what level of capitalisation counts as 'adequate', who audits the logs, who holds the shutdown power, and how an AIC would be taxed or would hold assets are all unanswered. It does not compare the plan with existing precedents such as Wyoming's DAO LLC statute, or explain how an agent-run company would meet federal know-your-customer, sanctions and securities rules, which a state sandbox cannot override. Norm AI's commercial interest as co-developer is not explored, and there is no response from Governor Meyer or from Zeberkiewicz. Only Comer's side of Congress is shown, with no Democratic or industry-group reaction, and there is no sign of which companies, if any, are lining up to use the sandbox.
