The Healey administration is turning to artificial intelligence behemoth OpenAI in a bid to make state government work faster and more efficiently, and pledged that its embrace of the controversial technology will come with “the highest standards for data privacy, security, and thoughtful, transparent usage of AI.”
A ChatGPT-powered AI assistant will be phased in across the almost 40,000-employee executive branch, starting with the Executive Office of Technology Services and Security, the administration announced Friday morning. Under a contract with OpenAI, Massachusetts is the first state with “enterprise-wide deployment of the tool,” Gov. Maura Healey’s office said. The state is also rolling out optional training programs to give employees using the AI assistant a baseline understanding of best practices.
“This is about making government faster, more efficient, and more effective for the people we serve,” Healey said. “AI has the potential to transform how government works, which is why we’re excited to partner with OpenAI on this AI Assistant, which will ensure a safe and secure environment for employees and improve their ability to deliver better service to the people of Massachusetts.”
The National Association of Government Employees, which represents about 15,000 state employees in Massachusetts, said it was unhappy with Healey’s announcement and said her administration is “rushing” the introduction of AI. A union official said NAGE “demanded to bargain” over the rollout when it was given a brief overview of the plan earlier this week and said some state workers have already expressed fears that AI could take away significant job duties.
On Beacon Hill, elected officials have been weighing the potential risks of AI against the economic upsides of a fast-growing industry. Influential lawmakers have indicated the House and Senate are unlikely to take up sweeping legislation regulating the technology in the near term, and business leaders have cautioned Beacon Hill against “creating unnecessary burdens” for the industry. Legislative action has been more precise so far. This week, the House passed legislation putting limits and disclosure rules on the use of AI in electioneering.
The administration said ChatGPT’s rollout will happen “within a walled-off, secure environment that protects state data and ensures that employee chat inputs do not train public AI models.” The state’s use of AI is governed by terms and conditions set by the EOTSS Privacy Office and other policies that govern the use and development of AI.
“By making ChatGPT available to the state workforce, we are empowering our employees with a secure, governed tool that can enhance service delivery while maintaining the highest standards for data privacy, security, and thoughtful, transparent usage of AI,” Technology Services and Security Secretary Jason Snyder said. “Our focus is not just adopting AI, but doing so in a way that reflects our values, and strengthens trust with the residents we serve.”
Governing in a state with deep roots in the tech industry, Healey has embraced AI, both in government and in her personal life. She and lawmakers included $100 million in a 2024 economic development law to create a Massachusetts AI Hub, and Healey said last summer that she is “less focused on regulation of AI” and more focused on “leaning into the investments that we’re making here on AI.”
At a Boston Globe summit last year, Healey said she had used a generative AI chatbot like ChatGPT to write a speech in her style, and to help her stepson with an assignment.
“I embrace AI, I think it’s amazing. I think it’s transformational,” she said.
With its procurement of an AI assistant for the executive branch, Healey’s administration made clear that it was looking to extend that embrace. In a request for quotes, the state said it was “seeking more than the procurement of a software tool.”
“We are looking for a partner who will work alongside us to make AI a lasting, responsible capability across government. Success will not come from technology alone but from close collaboration between the Commonwealth and its chosen partner. Together we will continue to build employee confidence through training and skill-building, help agencies identify and scale high-value use cases, and create a culture where responsible AI contributes significantly and is widely adopted.”
The state’s request contemplated an initial contract term of two years with an eye towards mutually-agreed extensions. Officials from the administration did not respond Friday to a request for a copy of the contract between the state and OpenAI.
Healey apparently announced the state’s plans Thursday evening at a kickoff party for the Massachusetts AI Coalition that featured a virtual appearance from OpenAI CEO Sam Altman, according to the Boston Globe.
The governor’s attendance at the event hosted by major tech sector businesses was not on her public schedule (her office did not issue a public schedule for Thursday). Her spokesperson did not respond when asked Friday why the governor’s office did not disclose her appearance at the party.
“Massachusetts is helping lead a broader shift across government to help public servants spend less of their time on paperwork and more of their time doing what led them to public service in the first place: serving the people of the Commonwealth,” Altman said, according to Healey’s office.
Bidding documents show that EOTSS declared Carahsoft Technology Corporation the apparent successful bidder on Jan. 27. Information submitted by Carahsoft, a Maryland-based government contractor that has other Massachusetts business and has reportedly worked to help OpenAI secure U.S. Department of Defense contracts, said OpenAI’s best and final offer is that it would charge the state between $1.56 million and $3.36 million based on how many employees use the AI tool.
In FAQ documents posted online, the state gave more detail about how it envisions state employees using AI.
“It should be used to support your role within the Executive Branch—such as drafting documents, summarizing reports, or researching topics relevant to your work. It should not be used for personal matters or to generate sensitive content without proper review,” the state tells employees in one document. “The AI Assistant is a tool designed to assist you in your role, not change your job description. It helps reduce time spent on repetitive tasks so you can focus on higher-value work. Your judgment, experience, and decision-making remain essential.”
Theresa McGoldrick, national executive vice president at NAGE, told the State House News Service that the union got a “very high-level, nonspecific overview of this AI tool” in a meeting this week, and that the union agreed to allow the administration to conduct what was described as a voluntary pilot program with EOTSS employees before an eventual rollout across the executive branch.
But she said the union objected to an email that went out Friday to executive branch employees announcing the new AI tool and that the administration’s announcement “is very different than what we thought we were agreeing to.”
“AI, although it can be an excellent research tool and can be used in many ways, can be dangerous to state employees in terms of if they put in any information that could be considered a data breach” or similar violation, McGoldrick said.
She added, “There are a lot of concerns at the highest levels of labor across this country about AI taking people’s jobs away. AI as a tool to aid someone in their job is very different, but we always have to be wary when were letting technology in.”
She said the union doesn’t object to AI outright but wants to work with the administration to protect employees. Early feedback at a board meeting Friday was that state employees are not eager to embrace the new tool, she said.
“A lot of the things AI can do would take away job duties, significant job duties. I don’t think that’s the intent of this tool … but that’s what was raised in the meeting today, or that they would be forced to use it even though [the administration’s] saying it’s voluntary.”
NAGE and the administration are scheduled to meet next week to continue discussing the issue, McGoldrick said.
“It’s sort of like putting the cart before the horse … they’re saying they will bargain with us, but they’re rolling it out,” she said. She added, “Bargaining is supposed to come first, that’s what the law says, and they’re putting the cart before the horse by doing this and … they’re rushing it.”
The state tells residents in a separate FAQ that they will still be dealing with humans when they call, email or visit a state agency. Residents are also informed that AI “is not being used to make decisions about eligibility or services.”
Residents “may notice” faster response times, clearer communication, more consistent service delivery and easier access to information thanks to the state’s use of AI, the administration said.
The administration tells state employees that the AI tool is meant to “support your work, not to actively monitor it.” But, “basic system activity is logged for security, audit, and compliance purposes. These logs are only reviewed if needed through proper legal and administrative processes,” the administration added.
The state also cautions workers that they cannot fully trust the output from the ChatGPT assistant because “it’s not perfect.” Employees are warned that it is to be “expected” that the AI bot will provide outputs that they do not agree with or like.
“It may occasionally generate incorrect or incomplete information. That’s why human oversight is essential — you must always review and verify the information you put into ChatGPT, and the outputs it provides, before using them in official work,” the state FAQ says. “It supports your work, but you remain responsible for the final result.”
The administration’s FAQs also note that “public models, such as public versions of ChatGPT, Gemini and Claude, are prohibited on state-issued devices,” as is providing non-public state data to those platforms.
