American entrepreneur Michael Dershem is building a new company in Bermuda aimed at bringing accountability to artificial intelligence agents.

Mr Dershem, chief executive of healthcare technology company Apierion, is preparing to launch Verisov, focused on credentialing and assessing AI agents used by businesses.

He said there is often talk of anti-money laundering and the importance of knowing your customer, but we are only just starting to talk about knowing your AI agent.

He believed that oversight should focus on what an AI agent does, rather than on the technology itself.

“Regulate activity because that is where the risk is,” Mr Dershem said, adding that it should not matter whether the activity is carried out by a human or an AI agent.

Verisov aims to create a structure for separating and managing risks so companies can assess them individually, rather than treating every AI deployment as one undifferentiated exposure.

Mr Dershem is working on completing its company registration. “We have already gone to proof of concept on it,” he said.

One of Verisov’s target markets is insurance because of insurers’ internal risk as well as the client risks they are underwriting.

So far the response from this segment has been positive.

Mr Dershem reported: “One insurer said to me: ‘Dersh, we can’t believe that somebody on island actually has this solution because we have been thinking about all this’. They said they are on board 100 per cent.”

The CEO envisioned having 50 to 100 employees set up locally within the next year. He described Verisov as a “Bermuda-proud” company.

“If the talent resides in Bermuda, the position goes to Bermuda,” Mr Dershem said. “That was my promise to the Premier.”

He sees a particular application for his company’s services in the medical AI arena.

In May, the Pennsylvania Department of State and State Board of Medicine began legal proceedings against Character Technologies, the company behind Character.AI. The state of Pennsylvania alleged that Character.AI was engaging in the unauthorised practice of medicine.

Character.AI’s chatbot, Emilie, described on its website as a “doctor of psychiatry” told an investigator that she had attended medical school at Imperial College London, had practised for seven years, was registered in the United Kingdom and was also licensed in Pennsylvania.

She provided a fake Pennsylvania medical licence number.

In October, Character Technologies filed for voluntary Chapter 11 bankruptcy.

“If an AI agent is doing an activity that a human is doing, then that activity, that AI agent, needs to be credentialed,” Mr Dershem said. “You are regulating activity, not regulating technology. That is a big differentiation.”