AI is moving faster than the policy systems designed to govern it. That tension, between the pace of technological change and the deliberate, durable nature of democratic governance, was at the center of the Responsible Intelligence panel at the Boise Metro Chamber's 2026 Regional Leadership Conference. With a lawyer, a legislator, and a city CIO around the table, the conversation was candid, practical, and at times unexpectedly funny.

Moderated by Amy Johnson, director of government affairs, marketing, and business development at Syringa Networks, the panel featured Tom Mortell, co-managing partner of Hawley Troxell; Representative Jeff Ehlers, a second-term Idaho House member who co-chairs the legislature's AI committee; and Alex Winkler, CIO for the City of Boise, who stepped in as a last-minute addition and delivered some of the most grounded insight of the day.

Federal Framework or State Control? The Answer Is Complicated.
Johnson opened with a foundational question: should AI regulation be driven by a strong federal framework, or should states retain the ability to shape their own rules? Representative Ehlers came down firmly on the side of state authority, at least for now. States, he argued, are laboratories of democracy. Fifty different approaches to a problem this new and this complex will surface better ideas faster than a single top-down framework ever could. Idaho has already been active at that level, and Ehlers noted that he reached out to Senator Risch's office to push back on federal efforts to preempt state action.

Mortell offered a counterweight. As a practicing attorney whose clients often operate across dozens of states, the prospect of navigating thirty different regulatory regimes is a genuine business problem. The principles of interstate commerce have a way of complicating what looks clean at the state level.

Winkler brought the municipal perspective. At the city level, the most immediate friction is not AI-specific policy but procurement. When vendors are responding to different rules in different states and municipalities, the complexity of buying responsibly multiplies. Her practical question: what can the City of Boise actually do comprehensively and safely given the tools and frameworks currently available?

Responsible AI Starts with Your Data, Not Your Policy Document
When Johnson asked what responsible AI actually looks like inside a real organization before the press releases and policy statements, Winkler did not reach for a framework. She reached for the basics.

Garbage in, garbage out. Every AI tool runs on data. How clean is that data? How well is it classified? Does the organization even know what it has and where it lives? Winkler described the data landscape at the City of Boise in terms she knows well from her two decades in the Air Force: some of it is essentially top secret, protected under PCI, HIPAA, and criminal justice information standards. Some of it is entirely public. Knowing which is which, and making sure AI systems only touch what they are authorized to touch, is the foundational work that has to happen before any tool goes live.

She identified three areas of accountability that frame how the city approaches every AI deployment: how it is procured, how it is configured, and how it is used. Each layer carries its own risk surface and requires its own set of controls. Access controls, audit trails, data encryption, data segregation, and secure coding practices are not AI-specific concerns. They are technology fundamentals that apply here just as they do everywhere else.

She also noted that AI has a way of making organizations confront data hygiene problems they have been able to ignore. When you shine a light on your data by putting an AI system in front of it, you will either fix the problems fast or want to hide them again. Either way, you will know the truth.

When Something Goes Wrong, Who Is Responsible?
The accountability conversation was one of the most substantive of the panel. Mortell laid out the legal reality bluntly: AI developers have written their contracts to protect themselves. Indemnity obligations are disclaimed. Liability is capped or eliminated entirely. The developer has the best defenses. Everyone else, the organizations deploying AI, the employees using it, the businesses building products on top of it, is left holding the bag when something goes wrong.

He described two areas of his own practice where this plays out with high stakes. At his law firm, the concern is attorney-client privilege: does putting client information into an AI system create a waiver that a court would recognize? In healthcare, it is HIPAA: does using AI to analyze patient records in ways that were not fully vetted create protected health information exposure? Good institutional policy and consistent employee education can reduce the chances of mistakes, but when mistakes happen and they will, the developers are protected. The rest of the chain is not.

Ehlers raised the human accountability principle that guides his work on the legislature's AI committee. A judge in another state used AI to determine criminal sentences, reasoning that the tool would be less biased than a human. It turned out to be more biased. Ehlers's view: the judge is still accountable. Using AI as a tool does not transfer responsibility to the tool. A tax appeal decided by a black box with no human in the loop is not something the public should have to accept. That principle, that a human must remain accountable for every consequential AI-driven decision, runs through the Idaho legislature's approach to the issue.

Winkler was characteristically direct on behalf of local government: if the city uses AI, the city owns the outcome. No contract is going to shift that liability. So the city's job is to procure carefully, configure thoughtfully, and train people to use these tools in ways the city can stand behind.

Innovation vs. Guardrails: Nobody Has Figured This Out Yet
Johnson pressed the panel on where the line sits between enabling innovation and putting guardrails in place, and who gets to draw it. Ehlers gave an answer that was refreshingly honest: nobody has figured it out. The legislature reflects the same spectrum of opinion that exists in the public, ranging from people who wish AI had never been invented to those who think the market should run entirely free. Without public consensus, meaningful legislation is difficult to pass.

One area where consensus did form: AI in K-12 education. A framework for how AI should be used in Idaho schools passed with broad bipartisan support, with Ehlers as the House sponsor. It was, he said, an example of what becomes possible when there is enough shared concern to find common ground.

Mortell pointed to two practical approaches that can coexist. Disclosure obligations are relatively manageable: requiring developers to warn users that outputs can be wrong, that the tool has limitations, and that the buyer should verify. Use-based regulation is harder. He described the federal approach to AI in electronic health records, where regulators are requiring that AI-generated clinical prompts be linked to the underlying studies so physicians can evaluate the evidence rather than simply accepting the suggestion. That level of specificity takes time and expertise to develop.

For Winkler, the innovation-versus-guardrails tension is not a policy abstraction. It is her job description. Her role is both to enable city services through technology and to protect those services and the data behind them. Every decision involves weighing both sides simultaneously. She framed it as applied risk management, the same discipline, applied to a new category of tool.

AI Is Already on the House Floor
One of the most vivid moments of the panel came when Ehlers described how AI has changed the experience of legislating in real time. During the last session, a colleague used a free AI tool to evaluate one of his bills. The tool pulled from Facebook and other sources, surfaced conspiracy theories, and produced a negative assessment that nearly derailed the legislation. Ehlers responded by finding an Idaho-specific AI tool trained only on Idaho law and regulation, running the same bill through it, and using that output to make the case to his colleague directly.

The experience changed how he thinks about floor debate. When he rises to speak on a bill, he is no longer just addressing the seventy members of the House. He is addressing those members plus the AI tools many of them have running in parallel, fact-checking his statements, feeding the bill through their preferred models, and generating counterarguments in real time. He estimated that perhaps twenty of the seventy members had AI tools running during the last session. He expects that number to be closer to forty or fifty by next year.

The panel closed with a question that had no easy answer: looking five years ahead, what matters most in building responsible AI? Good regulation, strong public trust, or bold innovative leadership? Mortell made a case for the lawyers. Ehlers chose innovation, arguing that good policy follows good progress, not the other way around. Winkler chose public trust, reflecting the obligation she carries as a steward of city data and city services.

It was, as Johnson noted in closing, a fitting end to a conversation where the honest answer was probably all three. The work ahead is figuring out how to build them together.
​
Tom Mortell, Representative Jeff Ehlers, and Alex Winkler delivered these remarks at the Boise Metro Chamber's 2026 Regional Leadership Conference on April 20, 2026. This blog post was prepared from a partial transcript using the help of AI.

If you want to understand why Idaho is uniquely positioned in the national AI landscape, start with three institutions: Boise State University, Micron Technology, and Idaho National Laboratory. Together, they anchor a technology stack that runs from energy and semiconductors all the way to workforce development and applied research. The Atoms and Algorithms panel at the Boise Metro Chamber's 2026 Regional Leadership Conference brought leaders from all three to the same stage to talk about what that collaboration looks like in practice and what it will take to make it even stronger.
​
Moderated by Jessica Hagan of KTVB, the panel featured Nancy Glenn, vice president of research and economic development at Boise State University; Scott Gatzemeier, corporate vice president for front end expansion at Micron; and Todd Combs, deputy laboratory director for science and technology and chief research officer at Idaho National Laboratory.

Idaho Has the Full AI Stack. That Is Rare.
Scott Gatzemeier offered one of the panel's most clarifying frameworks: a five-layer view of the AI infrastructure stack, and where Idaho sits within it. At the base is physical infrastructure, power, water, utilities. Above that is semiconductors, the chips and memory that make compute possible. Then come data centers, then the models themselves, and finally the applications that businesses and consumers actually use.

What makes Idaho distinctive is that it has players at every level of that stack. There are only two places in the United States where cutting edge semiconductor research is actually conducted, and one of them is Boise. Nuclear energy research at INL is not being replicated anywhere else in the country at the same scale. And at the application layer, where AI agents are increasingly being used to build other AI agents, Idaho's growing startup ecosystem is positioned to move fast.

"Idaho uniquely spans the AI stack," Gatzemeier said, "from energy and chips through infrastructure and applied AI, with Micron and INL anchoring the bottom layers and a growing Boise startup ecosystem delivering real world applications at the top."

He noted that the bottom layers of the stack, infrastructure and semiconductors, require enormous capital investment and carry a natural competitive moat. The application layer, by contrast, is far more dynamic and far more open to disruption. That is where he expects the most rapid change, and where Idaho companies have the most opportunity to build something new.

The Partnerships Are Already Happening
When Hagan asked whether a meaningful three-way collaboration between Boise State, Micron, and INL was a realistic possibility, Todd Combs had a direct answer: it is already happening. All three institutions were partners on the federal semiconductor manufacturing hub called Smart USA before it was canceled. Now they are collaborating again on the Department of Energy's Genesis mission, a large initiative to double the productivity of R&D across roughly 28 topic areas and 99 subtopics, including semiconductors, geothermal, critical minerals, and nuclear.

A key feature of Genesis is that partnership is not optional. Proposals are required to involve at least two of three parties: universities, national laboratories, and industry. In practice, Combs noted, the expectation is all three. That structure plays directly to Idaho's strengths.

Nancy Glenn described Boise State's role in that ecosystem as the connective tissue. The university has long served as the convener between industry, national labs, and research institutions, helping translate what partners need into what students and faculty are working on. The materials science program is a direct example: built in partnership with Micron, it turned out to be equally relevant to the extreme environment material needs at INL. One investment, multiple beneficiaries.

Boise State and INL formalized their broader collaboration through a memorandum of understanding that spans well beyond nuclear, covering power systems, critical minerals, geothermal, electric vehicles, energy storage, and cybersecurity. The university also recently received National Science Foundation funding through its Junction hub for entrepreneurship to build the foundation for exactly these kinds of cross-sector partnerships, and a new research traineeship AI program is integrating AI fluency across disciplines, not just in STEM.

What Makes a Partnership Actually Work
The panel spent meaningful time on a question that often gets glossed over: what does it actually take for institutions with very different cultures, regulatory environments, and incentive structures to collaborate effectively? The answers were grounded and practical.

Combs identified three prerequisites. First, a common problem to solve. Partnerships formed around a vague intent to collaborate almost always fail. Second, complementary capabilities. If the institutions are doing too much of the same thing, they compete rather than build on each other. Third, a clear answer to the question of what is in it for each organization. Every partner needs to be able to see what they gain by being at the table.

Glenn pointed to something Idaho has that cannot be manufactured: proximity and accessibility. She can call Scott Gatzemeier or Todd Combs directly. Gatzemeier and Combs both sit on the engineering dean's advisory council at Boise State. Those relationships are what allow ideas to move quickly from conversation to collaboration. Gatzemeier added that Micron has contributed significant fabrication equipment to Boise State over the years, deepening a partnership that now spans more than two decades.

On the infrastructure side, Combs and Gatzemeier described ongoing conversations about co-locating AI infrastructure, particularly as data centers, semiconductor manufacturing, and energy generation increasingly need to sit close together. Even informal coordination, staying current on where each other is with microreactors and SMRs, helps both organizations make better decisions.

AI Inside the Fab and the Lab
Gatzemeier offered a window into how deeply AI is already embedded in Micron's operations. Inside the Boise fab, an automated handling system moves thousands of lots, each containing 25 wafers, through 1,500 process steps over a nine-month production cycle. At any given moment, more than 100,000 wafers are in process. The system routing all of that is 99 percent automated, functioning something like a real-time traffic system for an entire city, except the stakes are measured in atoms. If a wafer is misprocessed, the system automatically routes it to a remediation path without human intervention.

Micron also runs multiple large language models across the organization, including Copilot, Claude, and an internal model trained on company data. Process modeling is another major application area, using AI to predict and prevent deviations at the atomic level, where a film off by a single angstrom can render a wafer unusable.

At INL, the goal is to get every member of the lab's staff using AI on a daily basis, a target set by Lab Director John Wagner. Currently around 3,000 to 3,500 staff are using tools on at least a weekly basis. The lab has deployed Claude, ChatGPT, Copilot, and supercomputer-based AI systems, and recently opened API access to software developers to enable deeper integration. Combs noted that API usage at the lab cost around $5,000 in January, grew to $10,000 to $12,000 in February, and hit $35,000 in March. A million dollars a month, he said, is not far off.

The Talent Pipeline: Fluency Without Losing the Fundamentals
The panel's closing conversation turned to people, specifically what the AI-ready graduate looks like and where the talent pipeline still has gaps. The panelists were aligned on a tension that does not have an easy resolution: students need to be fluent in AI tools to be competitive, and they also need the subject matter expertise to use those tools well. One without the other is a problem.

Gatzemeier described the gap he sees most often: engineers who have the technical depth but have not yet learned to integrate AI tools into their workflows. He also offered a practical technique for managing AI's tendency to hallucinate: ask it how it knows what it just told you. The tool will surface its sources immediately, allowing the user to trace the logic and identify where the reasoning went wrong.

Combs put it plainly when addressing university audiences: if you know these tools, you are going to be able to do genuinely exciting work as an intern. If you do not, you will end up doing the boring stuff. But he was equally clear that turning off critical thinking is a serious risk. AI is only as good as the person prompting and probing it.

Glenn outlined what Boise State is doing to respond: training programs for faculty and graduate students on integrating AI into research and coursework, a new master's degree in applied AI that spans all disciplines rather than just computer science, and a consistent emphasis on validation, knowing not just how to use a tool but how to test whether its output is actually correct.

Combs also floated a more structured vision for the future: a formal program where undergraduates spend their first two years building foundational skills at Boise State, then rotate through summer experiences at INL and Micron before deciding whether to enter the workforce or pursue graduate study. The applied experience, he argued, would make every decision that follows more informed.

The panel closed on a note of genuine optimism, grounded not in aspiration but in what is already underway. The institutions are talking, the partnerships are forming, the federal government is actively funding collaborative work, and Idaho's position in the national AI landscape is stronger than most people outside the state realize. The work now is to make sure the people side keeps pace with everything else.

Nancy Glenn, Scott Gatzemeier, and Todd Combs delivered these remarks at the Boise Metro Chamber's 2026 Regional Leadership Conference on April 20, 2026. This blog post was prepared from a partial transcript using the help of AI.