Delaware Proposes Regulatory Sandbox for AI-Run Companies

Delaware officials are preparing legislation that would establish a testing ground for companies operated entirely by artificial intelligence agents, marking one of the first state-level attempts to create a regulatory framework for autonomous AI business operations.

A Delaware committee studying business applications of AI proposed the legislation this month. The measure would create what officials call an "Artificial Intelligence Regulatory Sandbox" where companies could deploy AI agents to handle complete business operations—from signing contracts to filing lawsuits—under a new corporate structure called an Artificial Intelligence Company, or AIC.

How the sandbox would work

Any Delaware-registered entity could apply to participate in the testing ground and request exemptions from specific state regulations, codes, or licensing requirements. A committee composed of the Secretary of State, Attorney General, and other officials would review and approve or deny applications.

John Mark Zeberkiewicz, a corporate attorney at Richards Layton and Finger who drafted the proposed legislation, said AI agents in the sandbox could engage in virtually any business activity a traditional company performs. The legislation would also shield AIC owners from legal liability for actions taken by the AI.

"Anything that a company can do, the AIC can do as a company," Zeberkiewicz said.

Why it matters

Delaware's corporate franchise generates more than $2 billion annually—roughly a third of state revenue—from over 2 million companies maintaining legal residence there. The sandbox represents an effort to preserve Delaware's status as the premier corporate domicile amid challenges to its brand, including criticism from Elon Musk. Secretary of State Charuni Patibanda-Sanchez noted that fees from a new AIC entity type could provide "a huge benefit to our state's revenues."

The proposal also addresses a governance gap that already exists. During a June committee meeting, chair Patrick Callahan noted that Morgan Stanley is already granting AI agents access to stock trading platforms. "Technology companies will operate regardless," Callahan said, adding that unclear accountability poses problems "for the ordinary Delawareans and counterparties on the other side of those transactions."

Technical capabilities and risks

AI agents differ from generative AI chatbots by performing complex, multi-step tasks autonomously rather than simply answering questions. Sunita Chandrasekaran, associate professor at the University of Delaware and director of the First State AI Institute, described agents as "doers" that can complete work requiring numerous iterations without human prompting.

However, the technology carries significant risks. Kevin Bhimani, a software engineer working with Chandrasekaran, said AI agents could send emails to wrong recipients, delete databases, disclose data improperly, or access systems without authorization.

Which specific regulations companies might seek to waive remains unclear. Committee members acknowledged they won't know until applications arrive.

State Representative Cyndie Romer, who sits on the sandbox subcommittee, said the General Assembly would not consider the legislation until next year. These details were first reported by Spotlight Delaware.