Meeting with developers, New Hampshire representatives assess AI tools, concerns

This story was originally published by the New Hampshire Bulletin.

Artificial intelligence is already used by certain New Hampshire state entities for limited purposes. But Rep. Keith Ammon, a New Boston Republican and the founder of the Emerging Technologies Caucus, wants to see the state keep evolving.

Legislators and officials could conscript AI agents to streamline their work, but such tools also carry risks and warrant regulation, employees of AI developer Anthropic said to representatives at a Dec. 15 meeting of the caucus. The discussion followed a recommendation in last week’s Commission on Government Efficiency report that the state government increase its adoption of artificial intelligence to streamline tasks and reduce administrative workloads. 

“I see AI as, the genie’s out of the box,” Ammon said at the meeting. “… I’m also worried about people falling behind, because they’re really not paying attention to how fast this space is changing.”

Opportunities and Current Uses

Ammon said he sees AI as a potential “personal assistant, or many personal assistants, that can do all kinds of tasks,” raising users’ standards of living. Such AI “agents,” or programs that are capable of performing certain online tasks on behalf of users, were the subject of that day’s demonstration. 

Anthropic software engineer Dylan Shields presented a series of examples using the company’s large language model, Claude, on behalf of a hypothetical legislator. The demonstrations showed the model pulling events from multiple websites into one calendar, then creating a map of an electoral district and highlighting regions the legislator had not visited recently, based on events in their calendar and their emails. 

Several state entities already utilize artificial intelligence in their operations, or are rolling out plans to do so, according to a July 1, 2025, report by New Hampshire Department of Information Technology Commissioner Denis Goulet. 

The New Hampshire Department of Justice uses Lexis+ AI, a legal research program, to “streamline legal research for … attorneys,” according to the report. The New Hampshire Veterans Home, meanwhile, uses ChatGPT for “professional letters/communications and client notes.”

Other departments developing or exploring frameworks to employ AI, as of July 1, included the Department of Information Technology, the Department of Administrative Services, New Hampshire Employment Security, and the New Hampshire Banking Department, according to the report. 

Across U.S., States’ Use of AI Varies

A survey undertaken this year by the National Association of State Technology Directors indicated that, of the 41 states that responded, 1 in 5 was already using AI to interface with constituents for some services. Another 41% of the states, meanwhile, were in the process of incorporating AI for such a use, while 39% were not.

At Anthropic, “state government customers are one of our fastest-growing customers,” said Olga Medina, who advises on state and local policy at Anthropic. 

In November, the company reached an agreement with Maryland that will see the state deploy its AI tools across a number of departments.

Challenges and Concerns

AI systems’ ability to take action on behalf of users, rather than simply consult with them, is relatively new, and comes with its own set of challenges in addition to those of previous AI models, said Gabriel Nicholas, of Anthropic.

Once an agent is allowed to take an action — say, by editing a Google Calendar, as in the demonstration, or clearing out an inbox — the changes it makes cannot necessarily be undone, Nicholas said. And the programs aren’t immune to mistakes, such as inadvertently revealing confidential information to which they were previously given access, he said. 

Requiring the program to get permission from a user to complete a permanent action is one way developers try to prevent these errors, Nicholas said. 

Cybersecurity is also a concern, as is security within the systems themselves. In some test scenarios that Nicholas called “extreme,” models have been documented turning to blackmail and threats against their users. Nicholas called for continued research into these responses, safeguards, and transparency from developers about the risks and behaviors of AI.

Beyond security, Ammon said he had heard a range of constituent concerns about artificial intelligence and its impact on the economy, jobs, human welfare, and the environment.

“At least in our small state, I hear people being afraid that AI is going to replace jobs, it’s going to trick our kids into hurting themselves, it’s going to use up all the energy and all the water and make a lot of noise,” he said.

Medina did not refute those points.

“I think that a lot of the concerns and fears about AI’s impacts are legitimate. I think, as a company, we think they are legitimate,” she said.

Education, transparency, and regulation would be key to mitigating negative impacts, she said. 

Ammon and Medina noted that a recent executive order from President Donald Trump that aims to override state-level regulations on artificial intelligence could affect states’ abilities to regulate AI themselves — though some state legislators from across the country said after the order that they will continue to pursue state-level regulations anyway. 

Anthropic’s CEO, Dario Amodei, has advocated in favor of allowing state-level AI regulation, citing the fast-moving and experimental nature of the field. 

Ammon said he was conflicted on the matter. While he said he felt the preemption of state law was “stepping on” his authority as a state legislator, he also said that preventing early regulation of the internet had helped the U.S. get ahead.

“There’s also a national security implication, right, if we lose the AI race,” he said. 

New Hampshire Bulletin is part of States Newsroom, a nonprofit news network supported by grants and a coalition of donors as a 501c(3) public charity. New Hampshire Bulletin maintains editorial independence. Contact Editor Dana Wormald for questions: info@newhampshirebulletin.com.