Delaware plans test legal status for AI-run companies
A proposed artificial intelligence company structure would let AI agents manage entities inside a Delaware regulatory sandbox.
By Maya Lindqvist · Senior Technology Correspondent
3 min read
Delaware is proposing a new legal entity for businesses managed day to day by AI agents, according to Delaware Secretary of State Ms. Patibanda-Sanchez and Norm Ai founder and CEO Mr. Nay. The plan would test whether autonomous software can operate inside a defined legal structure rather than outside ordinary rules for liability, disclosure and oversight.
In a Fortune commentary, Nay and Patibanda-Sanchez said Delaware is working with Norm Ai on a framework for an artificial intelligence company, or AIC. They described the effort as a public-private partnership led by Norm Ai and said Delaware plans to test the model through a regulatory sandbox.
The proposed AIC would be a separate legal entity whose regular operations are directed by an AI agent instead of a human manager, the authors said. At the agent’s direction, the entity could sue or be sued, own and dispose of property, and take on obligations.
Each AIC would have one member, either a person or another entity, responsible for keeping it adequately capitalized, Nay and Patibanda-Sanchez wrote. The AIC would also have to keep a record of its activities.
The authors said that member would generally be protected from the AIC’s debts. They said the shield would not apply if the member failed to provide adequate capital or used the AIC to commit fraud or a willful legal violation.
Sandbox limits
Nay and Patibanda-Sanchez said AICs would be allowed to operate only inside a regulatory sandbox. Admission would be decided by a committee that includes the Delaware secretary of state, the state attorney general, the chief justice of the Delaware Supreme Court, the chair of Delaware’s AI Commission, and outside lawyers and technologists.
Under the proposal described by the authors, each AIC would have to meet capitalization rules and tell counterparties several facts before doing business. Those disclosures would include that the AIC is an authorized test entity, that Delaware does not endorse it, when the test period ends and how to file a complaint.
Officials would be able to suspend an AIC, revoke its authorization or ask the Court of Chancery to dissolve it, the authors said. Banking would be excluded from the program.
The sandbox would expire after 30 months, according to Nay and Patibanda-Sanchez. They said the sunset would give Delaware lawmakers a record to review before deciding whether to write broader legislation.
Why Delaware is pursuing the model
Nay and Patibanda-Sanchez argued that efforts to ban autonomous commercial activity are unlikely to succeed as AI systems become more capable. They said putting AI agents inside a recognized entity would make their actions easier for law, courts and counterparties to identify.
The authors compared the AIC to earlier legal structures that were debated when introduced and later adopted, including the modern corporation, the series LLC and the public benefit corporation. They said Delaware’s experience with entity formation and governance gives the state a basis to set rules for AI agents.
Nay previously co-authored a 2023 essay in Science that argued some state laws did not clearly prevent an AI agent from operating a company without a human in charge, according to the commentary. Nay and Patibanda-Sanchez said autonomous commerce is now possible with current technology and warned that activity could move offshore or onto anonymous systems if U.S. law does not provide a controlled venue.
They said consumer-protection and criminal laws would continue to apply to AICs. The aim, as they described it, is to test agent-led commerce under supervision, with capital linked to liability and a mechanism to shut down a system that breaks the rules.
This story draws on original reporting from Fortune.