Axios AI+ Government

October 17, 2025
Happy Friday! If you work in government and have a good story about how you're using AI, hit reply and tell us about it.
🗓️ Situational awareness: The AI event originally scheduled for next week with Treasury Secretary Scott Bessent keynoting has been postponed due to the shutdown, a source familiar tells Maria.
Today's newsletter is 1,480 words, a 5-minute read.
1 big thing: Trump firings could derail AI action plan
The administration's slashing of the federal workforce amid the government shutdown is threatening AI work at the Commerce Department, according to sources close to the agency.
Why it matters: President Trump has lofty goals for AI, but sources say they will be difficult to reach if the agency loses institutional knowledge and the staff needed to implement ambitious executive orders.
State of play: Employees across the National Institute of Standards and Technology, the Bureau of Industry and Security and the U.S. Patent and Trademark Office have been let go, though judge temporarily blocked the firings in a ruling this week.
- About 600 people have been laid off across the Commerce Department since Oct. 1, an Office of Management and Budget filing notes.
- A Commerce Department official told Axios that figure is closer to 440 people, but could not explain the discrepancy with the OMB number.
- These agencies are at the center of work to carry out the White House AI action plan. They are responsible for setting standards and reforming the way procurement works.
- The administration may be looking to replace some technical processes within the department with AI itself, according to sources, who spoke on the condition of anonymity to discuss internal plans.
Behind the scenes: BIS staff are affected, including personnel in charge of doing outreach to Silicon Valley and answering compliance questions, per sources familiar.
- At NIST, there have been fewer than 40 layoffs hitting the Manufacturing Extension Partnership and the Baldrige Performance Excellence programs, both proposed for elimination in Trump's budget request, sources said.
- The MEP Program is an advanced technology and cyber resource for manufacturers. The Baldrige Program is a federal initiative focused on helping businesses stay competitive.
What they're saying: "If you don't have the advanced technology of small manufacturers, then you're obviously not going to be able to compete with China," said Carrie Hines, president and CEO of trade association American Small Manufacturers Coalition.
- "You can't just come and bring AI to a manufacturer if they don't have the workforce to support it or the machines to implement it."
Per one industry source who works closely with Commerce, there is a belief that any cuts will be permanent, with technical roles replaced by contractors instead of rebuilding in-house capacity.
- Massive cuts to teams will derail progress on AI work, and the staffers that are left will struggle to keep daily functions going, the source said.
Catch up quick: Commerce tech agencies faced DOGE cuts earlier this year.
What's next: Commerce faces a deadline next week to implement an AI executive order, with companies gearing up to offer their proposals for the infrastructure, models and tools they want the government to designate as "priority" AI export packages.
- Public comments to the Office of Science and Technology Policy on federal regulations that hold back the development and deployment of AI are due later this month.
2. California embarks on its next AI chatbots fight
Gov. Gavin Newsom signed a flurry of California tech regulation bills, but it's the legislation he vetoed this week that could shape the national debate over online safety for kids.
Why it matters: The governor's decision underscores the tension between protecting minors and encouraging AI innovation.
What's inside: Assembly member Rebecca Bauer-Kahan's Leading Ethical AI Development for Kids Act would have banned "emotionally manipulative" chatbots, social scoring systems, and some facial recognition tech for kids.
- Developers would have had to classify systems based on the potential harm to kids, with high-risk systems facing stricter safety requirements, if it had been signed into law.
- Parents would have had to give affirmative, written consent before a child's personal information could be used to train a model.
State of play: Instead, Newsom signed a chatbot bill from state Sen. Steve Padilla earlier this week. That new law requires platforms to notify minors every three hours to "take a break" and that the chatbot isn't human.
Friction point: Some child safety advocacy groups viewed the Padilla bill as weaker than the Bauer-Kahan bill because protections for minors hinge on the platforms having "actual knowledge" that the user is a minor.
- That is a difficult standard for companies to obtain, making it easier for companies to claim they don't know who is a minor on their sites.
- Padilla's legislation also only requires platforms to disclose AI is being used if a "reasonable person" would be misled to think they are chatting with a human, which advocates say invites disputes over a vague definition.
3. How more states are making their own AI rules
We're back with part two of our state AI series featuring a peek behind the curtain with how Michigan, Maryland and Utah are using AI and thinking about guidelines around the technology.
Why it matters: States have a unique flexibility to test, deploy and set their own rules for AI as the federal government trails behind on legislation and regulation.
- How states procure and use the technology can set examples for others and inspire similar moves. Last week, we looked at the ways that Vermont, New Jersey, North Dakota and Pennsylvania are forging ahead on AI.
This is what we found out this week from state officials:
Michigan: The state is working with AI company Syncurrent to connect people applying for state and federal grants to relevant opportunities, said Jonathan Smith, senior chief deputy director of Michigan's Department of Labor and Economic Opportunity.
- His department partnered with Syncurrent and the Michigan Municipal League, a statewide association for cities and townships, to create the "MI Funding Hub" for local and tribal governments for things like affordable housing and infrastructure.
- "We reduce the time it takes local governments to find funding from a months-long process down to a few minutes," said Syncurrent cofounder Dhruv Patel.
- "We've helped over 100 communities identify grant opportunities," Smith said. "And this is at a time when there's way less volume of federal grant opportunities…The fact we're seeing usage go up, is really significant."
Maryland: Agencies are using Google Gemini, Microsoft Copilot and Amazon Bedrock to summarize and draft documents and do research.
- Maryland's Department of Human Services offers an AI chatbot for people to ask questions in English and Spanish to assist them with navigating the SUN Bucks program, which helped over 600,000 students this year afford food over summer break when school meals weren't available.
- The state is experimenting with how AI can make licensing and permitting more efficient. "If you want to start a business or get a tidal wetlands permit and you need to understand what the next steps would be, that's the kind of place where a chatbot would potentially be really helpful," said Maryland senior advisor for responsible AI Nishant Shah.
Utah: Utah has rolled out Gemini to most state employees, and its Commerce Department is using AI to process professional international licenses, like in nursing, for state credentials.
- Utah chief information officer Alan Fuller said it was an "insurmountable problem" to process licenses in different languages before staffers were able to use AI to help them translate and summarize the troves of information.
- Call center agents for the Utah tax commission are grappling with demands of a growing population, and the state is exploring how chatbots could help field questions.
- Utah is also developing a one-stop-shop online portal with an AI chatbot to route residents to easily do things like renew a license at the DMV or access social services.
4. San José seeks a custom generative AI platform
San José released a request for proposals to develop a custom generative AI platform for its more than 7,000 city employees, Mayor Matt Mahan's office announced.
Why it matters: San José is all-in on using AI to transform city services.
- The RFP today follows an AI training initiative for city employees.
- San José is an example of one city that is diving headfirst into using AI, while others are experimenting more slowly.
What they're saying: "We're working to integrate generative AI into everything we do at City Hall because you deserve government that is fast, responsive, personalized and cost-effective," Mahan said in a release.
- "Thanks to our workforce up-skilling curriculum, we're already seeing city workers deploy these new tools to automate routine aspects of their jobs and deliver more impact for our community."
How it works: City employees want to use the custom platform to analyze and summarize documents, draft reports, review code and generate charts, per the release.
In our recurring feature "Using AI," we note use cases ranging from light to serious on how governments and bureaucrats are putting artificial intelligence to work.
Thanks to Mackenzie Weinger and David Nather for editing and Anjelica Tan for copy editing.
Sign up for Axios AI+ Government






