Why should Delaware care?
Artificial intelligence forms the backbone of the hottest industry in the world right now. Its use also has sparked questions about whether the technology can be utilized safely and for the benefit of society. Delaware officials now say they may create a testing ground to allow AI agents to run whole companies.
A Delaware committee that has been studying the business uses of artificial intelligence proposed legislation earlier this month to temporarily ease state regulations on companies deploying the fast-growing technology.
The proposed legislation would create a testing ground for companies to use what are called AI agents to autonomously complete business tasks typically done by humans. The AI agents would oversee whole business operations under the umbrella of a new kind of entity, called an Artificial Intelligence Company, or AIC.
Supporters say the resulting regulatory “sandbox” would allow Delaware to test how autonomous AI businesses operate, and provide lawmakers with key data to develop rules governing their use within the state’s prominent corporate franchise.
The legislation is likely to be introduced in the General Assembly next year.
Its drafting this month emerged after tech companies plowed billions of dollars in recent years into the AI industry — much of it to build data centers across the country to power the technology.
The proposed legislation also comes just as Delaware seeks to retain its status as the preeminent legal home for companies, even in the wake of attacks on its brand from the world’s richest man, Elon Musk, and others.
The principal drafter of the proposed legislation, John Mark Zeberkiewicz, said the measure could allow AI agents to engage in just about any business activity — from providing coding services to signing contracts, or even filing and defending lawsuits.
He also noted that it seeks to protect owners of new Artificial Intelligence Companies from facing legal liability from actions the AI might take.
“Anything that a company can do, the AIC can do as a company,” said Zeberkiewicz, a corporate attorney at Wilmington’s Richards Layton and Finger.
But exactly how an AI might work in the testing ground remains unclear – reflecting an larger uncertainty surrounding the broader AI industry.
Asked whether an AI company could take on whole operations of even risky businesses, such as a small nuclear power plant, Zeberkiewicz chuckled, then said “fair enough.”
He said he does not believe that is something a company would do in the real world.
Supporters of the sandbox argue that Delaware cannot wait for such questions to be completely answered because businesses are already deploying AI agents to perform autonomous work. They say the legislation would allow the state to develop legal standards before the technology becomes more widespread.
During a June meeting of the Delaware committee that studied the issue over the past year, committee chair Patrick Callahan stated that the investment bank Morgan Stanley is already granting AI agents access to stock trading platforms.
He said the future the committee is planning for “is actually here right now, and it’s arrived through one of the largest financial institutions.”
While AI agents are already transacting, Callahan said there is no clear party being held accountable when things go wrong.
“Technology companies will operate regardless. It is a problem for the ordinary Delawareans and counterparties on the other side of those transactions,” he said.
It is not immediately clear whether any large company or lobbying group asked Delaware to propose its AI testing ground. Still, in April, the prominent billionaire Mark Cuban posted on the social media site, X, that a U.S. state should allow anyone to form a company using an AI agent and a stablecoin digital currency.
In response to the post, Gov. Matt Meyer stated “On it!” He then provide a link to Delaware’s AI sandbox initiative.
Who’s liable?
The incentive for a company to enter into the Delaware’s proposed regulatory sandbox would be to test an autonomous entity with a liability shield, Zeberkiewicz said.
“It’s like any limited liability company – you form it for the purpose of making sure that the owners of the business are not automatically liable for the debts and obligations of the entity,” he said.
Any entity formed in Delaware could apply to be a part of the testing ground, which would officially be called an “Artificial Intelligence Regulatory Sandbox.” In the application, the company could ask a future oversight committee to be exempted from any Delaware regulation, code provision, or licensing agreement.
The future committee — which would be composed of the Delaware Secretary of State, the Attorney General, and others — would then accept or deny the applications for the testing ground.
Allan Kyle, a member of Delaware’s AI committee, said he can’t foresee which regulations applicants might ask to be exempted. Callahan said the same.
“We’ll find out when people apply and say, ‘Hey, we’re being prevented from doing commerce because of this one Delaware regulation that we’re required to do and we want to look for a waiver for that,’ “ Callahan said.
State Representative Cyndie Romer, who sits on the sandbox subcommittee, told Spotlight Delaware the proposal would not be considered by the general assembly until next year.
What is an AI agent?
Generative AI–which includes big name chatbots like OpenAI’s ChatGPT, Anthropic’s Claude, and Google Gemini – is typically what most people think of when they hear AI.
Sunita Chandrasekaran, associate professor of computer and information sciences at the University of Delaware and director of the First State AI Institute, said that for larger and more complex tasks, generative AI would need to be prompted numerous times and require many “iterative back-and-forths”.
“You won’t get what you want in the first iteration,” Chandrasekaran said.
Chandrasekaran said an AI agent is the “doer,” while generative AI answers questions.
“You’re asking more than a question – you’re asking it to do a job for you,” Chandrasekaran said.
Among Chandrasekaran’s ongoing research is a project to build AI agents to write code. The agents can generate about 10,000 codes in a matter of a few minutes, which Chandrasekaran said she could not do manually.
“It would take me maybe a couple of years to write that many codes,” Chandrasekaran said. “The agent is now giving me more output than a human can ever possibly do in a given point of time.”
Many companies have announced layoffs related to new AI efficiencies. Some specifically cited new AI agent programs that fill in for customer service, support engineers, and management jobs.
But along with efficiencies, AI agents have also introduced new risks for businesses.
Kevin Bhimani, a software engineer on Chandrasekaran’s team, said the risks could range from sending an email to the wrong person to deleting a whole database or disclosing data to the wrong people.
“It could hack into systems and access data that it wasn’t meant to access,” Bhimani said.
Why in Delaware?
During a Delaware AI Commission meeting this month, Delaware Secretary of State Charuni Patibanda-Sanchez said the program would allow Delaware to be on the “cutting edge of technology,” if the testing goes well.
And, just like the state charges fees for LLC formations, Patibanda-Sanchez said Delaware would benefit from fees paid to form Artificial Intelligence Companies.
Currently, the collection of taxes and fees on the more than 2 million companies that maintain their legal home in Delaware contributes more than $2 billion to the state budget, or about a third of all revenue.
“Should a new entity form be created out of it, that’s really a huge benefit to our state’s revenues,” Patibanda-Sanchez said during the June meeting.
Patibanda-Sanchez said “we want to be able to exit the AIC Act out of the sandbox” and into state’s laws, if AI agents prove to be a beneficial and responsible use.
“That’s the whole reason to test it, instead of just trying it out for the first time and not knowing what’s going to happen,” Patibanda-Sanchez said.
