S.B. 322 Educational Technology Regulatory Sandbox: transcript

Bill. I would love to. I would love to thank you. SB 3 22. I have sat, you know, in, in your chair over in the Senate and to a lot of bills that came through this year, uh, about ai, about software in the classroom. And I, I understand the intent behind all of them. I, I think it's really important that we protect children. This is back to, to that same theme, but I, I also think is, is it is a mistake if we throw out the baby with the bath water. Uh, I think there are legitimate concerns about ai. I I think these concerns are important, but I think what we need is guardrails, not complete blocks. Artificial intelligence is already in Utah classrooms. Teachers are using it, students are using it, vendors are selling it. But right now, Utah has no systematic framework for evaluating whether any of it works, whether it is safe or whether it respects our students' privacy and our parents' rights. Without a structured framework, we face three risks simultaneously. First, risk districts adopt AI or software tools with no safety vetting, exposing students to inappropriate content, data harvesting or systems that simulate emotional relationship with minors. Legislators react to bad actors with blunt prohibitions, shutting out genuine innovation that could help struggling readers, English learners, and students with disabilities. Risk three, statewide adoption happens through administrative mandate rather than legislative authorization. Bypassing the deliberative process. This chamber exists to provide. This bill creates Utah's educational technology regulation sandbox a structured, evidence-based, evidence-driven framework for evaluating AI in public education before any statewide commitment is made. It is not a ban, it is not a blank check. It is a discipline process. A time limited pilot program. No more than three years, allows schools to test AI tools under real classroom conditions while the state monitors the results. Participation is entirely voluntary for schools, educators, parents, and students. Before any AI system can participate in a pilot, vendors must complete red teaming. Red teaming is adversarial testing specifically designed to expose what happens when a minor tries to manipulate the system. So if a minor tries to break out of the con confines of the system, that the whole idea of about red teaming is let's build in protections that actually detect this. And this is something that's possible with AI that we haven't had in existing systems. I, I hear testimony of, of kids that you, you know, we had one in our, our committee where, where a child testified how he was able to break out of the safety protocol and go anywhere he wanted on the internet. This red teaming, uh, w will actually help with that. Parental authority is preserved. Parents receive written notice before any AI interaction with their child. They can opt out at any time. In fact, we've got, we, we we're working on a, uh, a substitute that actually will, will prevent, uh, will, will actually make it so that parents actually have to opt in to start with, and then they can opt out anytime. No AI system under this bill can assign final grades, make placement decisions, or override a teacher's professional judgment without human review and approval. The human in the loop requirement is not a suggestion, it is a condition of pilot authorization. Educators must review and appro approve before any AI output affects a student's official record. And if the state board of, uh, identifies a violation, the bill requires documented evidence, written notice to all parties within 24 hours. This is perhaps the most important structural feature of the bo of, of the bill. No AI system can be adopted statewide without explicit legislative authorization. That doesn't mean it can't be used in individual districts, it just means that we can't adopt it statewide until it has gone through this process. This bill does not merely protect students from bad technology. It establishes a student due process and dignity framework. AI systems are prohibited from simulating robotic or personal relationship with students. Crisis response is integrated with safe ut If a student expresses self-harm ideation, the system must immediately connect them to crisis resources. Student data cannot be sold used for advertising or retained after pilot completion without parent authorization. It creates a safe harbor for de-identified student data to enable responsible research and evaluation. While protecting individual Vendors own controlled or subject to the dur jurisdiction of a foreign adversary are prohibited from participating a national security safeguard that exceeds most states' language on students' rights. Students have the right to know that they are interacting with an AI rather than a human. Students can request explanation of how any AI generated recommendation or assessment was produced. Students can request human educated review of AI generated decisions before it takes final effect. Students can contest AI impacted decisions through existing school appeal procedures. Students cannot be penalized for declining optional AI tools or requesting alternative assessment methods. This bill does not ask us to trust that AI works. It requires proof. And so independent a evaluators before any statewide, uh, adoption will have to, uh, do pre and post tests on the students to make sure that, that these systems are actually useful, that they're actually producing results. This driven framework actually will be important for us. The offices of regulatory relief would admi administer pilot applications and maintains the Public Regi Registry of Approved Pilots, the Safety Enforcements State Board of Education and enforces student safety and privacy requirements and can suspend or terminate Pilot Education Technology Advisory Council advises on student standards evaluation methodology and red team adequacy. The state AI coordination Office of artificial intelligence policy coordinates to avoid duplicative requirements and align with broader state AI policy. The legislature authorizes any statewide adoption based on evidence compiled from com complete pilots. The liability here is carefully balanced. These, these what what specified or actually, uh, are, are actually guardrails and not prohibitions. And, and, you know, these guardrails can actually change, uh, the, uh, the board will be able to, uh, update them. There will be, uh, a five-year Sunset on this. So why this Utah has consistently led the nation in technology policy that balances innovation and protections. We created the first state office of artificial intelligence PO policy. We have established the strongest student data privacy framework in the region. We just passed another rule on privacy, uh, propagated by the state board in a, in a bill that has already passed. Uh, we have demonstrated repeatedly that we can write AI policy that respects both the promise of new technology and the cons, constitutional liberties of people it affects. The bill also provides technical assistance to Utah based ed tech companies helping local innovation meet high safety standards. I'm happy to answer any questions.

Thank you Mr. Chair. Now, this may sound like an obvious question, but how do you define te Educational Technology Regulatory Sandbox?

So what the sandbox, the, the idea behind a sandbox is it says, look, we want innovation. And so we're gonna block off and we're gonna take out the regulatory re regime, but we wanna retain these guardrails. Okay? And so you, you know, they can do innovative things. They don't have to follow the same pedagogy standards that we've had in the past, right? But they have to prove that what they're doing actually is gonna make students better off. But the sandbox has actually created so that we can have innovation.

Can I, so that we can try new things. Can I follow up please? Yes, please. Um, so the next, uh, online tenant says, establishes minimum standards. Why would you just want minimum standards for

Artificial intelligence? Well, the standards would be the standards on safety. Okay. That's the guardrails that I'm talking about, right? And, and those are actually list listed specifically. These are the things that every, every one of the bills I've seen this es this, this session has said, look, here's, here are the problems with ai. But the conclusion often is, let's just block it off and not allow any of it. And I think that's a mistake because AI's here, AI is gonna be an important part of education going forward, but we have to make sure that we keep teachers in the loop that we, that we put up guardrails so students aren't harmed.

Thank you. What a lot of my texts e from parents are saying, um, first of all, they don't want their kids in front of Chromebooks all the time. And I hear that all the time from students in my area. So it sounds, and they want their kids to be protected. So it sounds more like they don't want to be using this, but they do have to have consent, they have to opt in. Is that correct?

Yes, that's true. To do this. Okay. Because there is I think, some misunderstanding between what this would do and what their kids would then be doing. Yeah. And, and what I would hate to do, I, I think there's a lot of good, uh, we, we've seen, uh, reading software that's really been handy in helping students learn to read. Uh, there's a lot of good software out there, and I, I think, you know, we have to move away from this idea that the sky's falling and it, you know, everybody's gonna die because, you know, some AI bot's gonna take over the world and it's here. Right? So responsibly, we need to put up guardrails for it. But, but I don't, I don't believe we really want to ban it entirely. Just like I, I don't, you know, if, if we start evaluating tech this way, I think we could solve a lot of the concerns and the fact that this is opt-in, right? Thi this may be a charter school that picks this up. This may be an accelerated class at a, at a, at a public school that picks this up. But I, I think we've, I, I think we've, you know, we need to evaluate this based on data at, at the end, right? We have to show that this actually gives us better performance.

Thank you, Mr. Chair. Um, I would like to make a motion that we favorably recommend, uh, first substitute Senate bill 3 22.

Like I said, I think Utah has consistently led the nation in technology policy that balances innovation with protection. And, and I think we, we would make a, a serious mistake if we simply ban this technology, I, or, or completely remove technology from the classrooms. We need to be responsible about it. We need to have guardrails. And, and I think, uh, I think this, you know, I I actually sat down and wrote this bill because I, you know, I, I don't know if you guys know, but I, I've, I actually built my career around AI and technology and, uh, did, did quite well with that. Uh, I, I did a lot of early research on neural networks, genetic algorithms. In fact, my company built one of the first, you know, our, our application was essentially a large language model. It, it, it actually, uh, managed, uh, underwriting of appraisal documents. And I, I understand fundamentally how this works. These are predictive algorithms, okay? They don't always give us the, the, the answer we want, but I can tell you they're moving the world forward. And, and if we simply ban this stuff from our public schools, our kids are gonna be left behind. And that's why I think we have to have responsible frameworks. We have to have responsible guardrails, but, but we also need to find solutions that, that help us to better teach our kids. And with that, I, I would appreciate your support.

Uh, yeah, I'll just make one comment or a couple comments. Um, I like the idea of, of creating these guardrails. I do think that that is very, very important because I agree the ai, AI is here. We do need to have those guardrails. I am a huge proponent though, of limiting a lot of technology in school, um, simply because I think we're starting to see results that aren't what we thought they were going to be with having too much technology, especially in the younger ages. Um, so that is a huge concern, but I love the fact that it is opt in, right? It's, it's, they don't have to use it. People don't, they don't, they're not, um, forced to use AI in their schools. I love the Loa local control component. And so from that perspective, I think it's important that we do have, for those that are going to use it, we do need to have some guardrails because it could get outta control for sure. So I appreciate, uh, the good Senator Cullimore. Senator Cullimore, bringing this before us.

Um, I do not. You tell me. We'll

Oh, first and this substitute need to be moved. Yes. If we could do That all to entertain a motion. I'll move that. Okay.

I, I started writing this bill when I saw a lot of the, uh, AI bills that were coming forward this time, and a lot of the software bills. And so the new, uh, the new title actually is Education Technology Regulatory Sandbox. Because I, I, I actually think that we need to codify and, uh, explain some of the rules. I I heard from a lot of people, this is you. You know, I, I think we're trying to go from one extreme to the other, where we have software in school, we have, uh, machines in school, and then we move to the other extreme where we want to get rid of everything. And I, I think there's a balance in the middle. And I, I think the parents groups that have, have been in have some very, uh, valid critiques. But what I wanted to do, rather than just having vague notions, I actually wanted to articulate how we could actually satisfy a lot of those rules and have a framework whereby we can, uh, we can actually move stuff forward. Uh, you know, artificial intelligence is already in Utah classrooms. Teachers are using it, teachers are using technology, students are using it, vendors are selling it. But right now, Utah has no systematic framework for evaluating whether any of it works, whether it is safe, or whether it respects our students' privacy and our parents' rights. Without a structured framework, we face three risks simultaneously. First, districts adopt AI tools with no safety vetting, exposing students to inappropriate content, data harvesting and or systems that stimulate emotional relationship with minors. Legislators react to bad actors with blunt prohibition, shutting out genuine innovation that could help struggling readers, English learners, and students with disabilities. Statewide adoption happens through administrative mandate rather than legislative authorization, bypassing the deliberative process. This chamber exists to provide. This bill creates Utah's education, technology, regulatory sandbox, which a actually already exists. And that's what one of the things that this, this thing points out. We already have a regulatory sandbox for, for education. And I think this, this actually clarifies the relationship of the board, the relationship, uh, to, uh, to innovation. So what does this do? Uh, it's, it's not a ban, it's not a blank check. It is a discipline process. Time limited pilot programs, no more than three years, allow schools to test a EI tools under real classroom conditions, while the state monitors results. Participation is entirely voluntary for schools, educators, parents, and students. I, I want to emphasize that this is voluntary safety. Before students, an ongoing, before any AI system can participate in a pilot, vendors must complete red teaming adversarial testing. And so first we want proposals that come in. We want to vet those. But, but before these are de deployed, the, the, uh, vendors actually have to complete red teaming adversarial testing, specifically designed to expose what happens when a minor tries to manipulate the system. We require documentation of what was tested, who tested it, and what vulnerabilities were found. And fixed. No red teaming, no pilots. So this is not a one-time certification for multi, multi-year pilots. Vendors must conduct updated red teaming annually and report any safety failures to the state within 48 hours. The monitoring does not end at approval. It continues for the life of the pilot. Now, what is, what is red teaming? Red teaming is actually the software itself trying to break out. Uh, we heard in a testimony before that that, uh, of students that came up and said, yeah, well, you know, I broke through the safety protocols and I was able to access anything I wanted. We wanna make sure that that vendors are aware of that upfront. And one of the great things about AI is it actually will allow us to do that type of, of testing and allow us to build those things into the software. So educator training also is an important thing. And, and before using ai, I think educators not need to be trained. And many of our teachers already are very aware of software systems. They've been using it for a while. I don't want to eliminate those. I think what what this is, is kind of going forward. How are we going to deal with this, uh, more complex software? But, but we do need to train educators. I I think it's important we have, we have some very good educators that work with, uh, software systems already. I, I think, uh, training on these new, on this new tech is important. But I, but I do think that, you know, that we have to provide training in multiple formats online, in person and hybrid. So, uh, the first substitute actually, uh, strengthen this provision, making it explicit that educator training must be completed. Uh, before we have AI interaction with students, uh, parents parental authority is preserved. And so to, to begin with, people have to opt into this. But at any time, parents, parents will receive a written notice before any AI interaction with their child. They can opt out at any time without academic penalty. They can expect they can inspect AI generated records, request data deletion, and meet with educators about any concern. Their child's education does not depend on their consent to a, I use educators stay in charge. No AI system under the spill can assign final grades, make placement decisions, or override a teacher's professional judgment without human review and approval. The human in the loop requirement is not a suggestion, it is a condition of pilot authorization. Educators must review and approve before any AI output affects a student's official record. And if the state board identifies a violation, the bill requires documented evidence, written notice to all parties within 24 hours, and an opportunity to respond before any non-emergency enforcement action. This ensures oversight is real, but not arbitrary. So the, the, uh, last thing here is legislature re the legislature retains the final authority. So this is perhaps the most important structural feature in the board. No AI system can be adopted statewide without explicit legislative authorization. The State Board of Education reviews pilot evidence and makes a recommendation to this legislature. Uh, the board cannot mandate, cannot condition funding, and cannot require specific vendors without the legislature approval. And I think this is important because this is kind of a new frontier that we're, we're moving into. And so, uh, you know, let's look at student protections. It, uh, on safety and privacy. AI systems are prohibited from stimulate simulating romantic or personal relationship with students. Crisis response is integrated with safe ut if a student expresses self-harm ideation, the system must immediately connect them to crisis resources. Student data cannot be sold used for advertising or retained after pilot completion without parental authorization. And if you read details in the bill, it, it, basically, the student data can be integrated in the system as far as the types of things they do to break the system. And that's part of the what the re red teaming does. And I, I think that's really important, but it has to create, it creates a safe harbor for di identified student data to enable responsible research and evaluation while protecting individual privacies. So, so the vendors have to prove that this, any data that's integrated in the system cannot be de identified. Okay? So, so it, it can't be joined back together. And point to specific students, uh, vendors own control are subject to the jurisdiction of a foreign adversary, or prohibited from participation national security safeguard that exceeds most states language on students' rights. Students have the right to know that they are at interacting with an AI rather than a human teacher tutor or counselor. Students can request explanation of how any AI generated recommendation or assessment was produced. Students can request human educator review of any AI generated decision before it takes final effect. Students can contest AI impact decisions through existing school appeal procedures. Students cannot be penalized for declining optional AI tools or requesting alternative assessment methods. The, uh, the evaluation of the system that this bill does not ask us to trust that AI works. It requires proof. Independent evaluators selected competitively and free from vendor relationships must apply research standards that include quantitative pre and post assessment of student learning control or baseline comparisons to isolate AI's ax actual contribution, disaggregated subgroup analysis to identify any desperate impacts or on collected on protected student populations, effect size calculations to determine whether outcomes are not just statistically significant, but practically meaningful standards for eternal internal validity, external validity, reliability, and objectivity. This is not anecdotal, uh, driven policy. There, there are standards for how it does. So, you know, we have application and monitoring the reg, the Office of Regulatory Relief administers pilot applications and maintains the Public Registry of approved pilots Safety enforcement. The State Board of Education enforces student safety and privacy requirements and can suspend pilots for violation of state or federal laws subject to due process procedures. And these are with respect to the, the, uh, the student safety and privacy requirements. Educational Education Technology Advisory Council will advise on student standards evaluation methodology and red team red teaming adequacy. Uh, so the first substitute requires at least one ad hoc member with K through 12 classroom experience whenever education technology applications are under review. Uh, state AI coordination, office of Artificial intelligent policy, whose director is now a permanent voting member of the advisory council under the first substitute coordinates to avoid duplicative requirements and align with broader ai, uh, broader state AI policy and the legislature off authorizes any statewide adoption. And so, uh, you know, the liability is carefully balanced, gross, negligent, or willful misconduct does not apply the protections violation of student privacy law conduct outside the scope of the approved pilot. And so, um, that, that pretty much concludes my presentation. I I didn't want to go on too long because I know we have a full agenda. So, uh, with that, I'm open for questions. Thank You. Uh, we'll turn the time over to Senator Riebe.

Thank you, Madam Chair. So I, I like a lot in this bill. And so I just have one question that I wanna ask now and maybe some other questions later. But on line 6 7 6 77, it says, an alternative assessment here in the state, we have so much choice. We just heard about how much choice we have, and yet, every time we're in a brick and mortar neighborhood school, we have all these alternative assessments, alternative programs, alternative choices, alternative, um, reports. Um, this is the, this is the part that bothers me. Like an alternative assessment that has, doesn't have AI in it. Well, everything has AI in it. I mean, you literally type into your Google search bar and it interprets what you're saying, which is ai. You could even say that Grammarly is ai, Microsoft Office Word, uh, word choice is ai. So to what point would you have to determine that you didn't wanna have AI in your assessment? Because we don't really have that capability. So that's my sticking point with this right now.

I I think you're raising a very valid point. Thank you. I, I think that's, I think that's an important consideration and, uh, we could work on that language a little bit because it, you know, parents do have the right to opt out of these things, but, you know, that doesn't mean they get to take the class if they opt out of it. Right. They may have some other class that they could take instead. And so I, you know, I think as this goes on, I, I think people are gonna get a lot more, uh, used to it. And I think that you, you, you know, one of the things that that is pointed out here though, is the teachers make that final determination of, of grades and the assessment and other things, right? So, yeah, I, I, I agree with you. We do have to get rid of that kind of ones, uh, point.

Thank you. Uh, the reason, part of the reason why I like the spill is it is whiplash. We need continuity in what's happening with our technology. I think this provides a better continuous experience. But we have whiplash every year. We come up here, we make a ton of laws. We have whole entire educational communities coming up here begging us not to do stuff. And so when we think about how we're gonna move forward, we have to move forward with more than a one year plan and changing it on a dime every year. Uh, I know right now we have, we've been pushing for years to have devices available for students that wanted to use them this year. No one wants us using devices, it feels like. So I think that this provides a little bit more continuity. So thank you for having a five year pilot and, um, for coming up with some alternatives to just saying No. Thank you.

Yeah. I, I, Senator Riebe, I, I, I sat here and listened through every one of these presentations. And I, that's, that's why I started writing this, because I thought, you know, we do need guardrails, but, but you know, some of the first AI bills I saw said, let's, let's just ban it, right? Let's ban chatbots. Let's not allow students access to them. And, and the reality of it is, it's here, right? And, and you know, what we need to do is give it a chance. Let let people look at the data. Let's let, let's let schools adopt it. Let's let them move forward with pilot programs, and then let's evaluate them. Let's use a data-driven approach to, to evaluate. Uh, but we do have to make sure there's safety guardrails on it. And I, I think this lays out the, the most important ones. And I, and I think this is a very pro technology bill. Uh, and, and I think that's important moving forward. So thank you.

Corinne Johnson, um, president of Utah Parents United. And we are here in support of this legislation. We know that the majority of Utah schools and the majority of Utah students are all ready using AI in the classrooms. This legislation. Um, we, we would've loved it if we had been here three years ago, but we're here now, and it's very, very important that we're here talking about this space. We feel that this legislation fits beautifully in with the other bills that have been presented. And I did wanna comment, there has been no Bill presented this session that, that bans, um, chatbots in education. Um, but we do have the Balance Act, um, which talks about one-to-one devices. We have the Safe Act, which talks about contracts with all education software, and this is perfect to fit in to talk about AI and creating a sandbox. The one comment we have is Line 2 55, which states the participation is voluntary. It states it, but it doesn't provide a process for that, um, voluntary parental opt in. I love the opt out piece. It is so beautifully written and so clear that once you've opted in, you can opt back out. But we would love some clarification on the procedural expectations, um, for that voluntary participation by parents, um, school districts and others into the sandbox. And with that, we, um, um, are supportive. So thank you.

And would you like to some, Yeah, I would say, uh, Utah's consistently led the nation technology policy.

Thank you, Madam Chair. I'd like to send Senate Bill 3 22 AI and education grant amendments out with favorable recommendation. Thank first substitute. Thank you. It's a team effort.

You. I, you know, I, I guess the last thing I'd say, why this, why now, uh, Utah has consistently led the nation in technology policy that balances innovation and protection. We created the first state office of artificial intelligence policy. We have established the strongest student data privacy framework in that, in the region. We have demonstrated repeatedly that we can write AI policy that respects both the promise of new technology and the constitutional liberties of the people. It affects this bill. The, the bill also provides technical assistance to Utah based ed tech companies, helping local innovators meet high safety standards and complete compete nationally keeping economic benefits. Here in Utah, SB 3 22 is the next step in that tradition. It does not wait for harm to occur before responding. It builds the evidence base. The legislature needs to make sound decisions about a technology that will shape how an entire generation of Utahans learns to think, to work, and to engage as citizens. If we fail to build structure now, structure will be imposed after harm occurs. This bill allows Utah to lead deliberately, transparently, and constitutionally. Thank you very much. I, I, I hope I can get your support.

Not someday, not in theory right now. Teachers are using it, students are using it. Vendors are marketing it aggressively to every school district in the state and across the country. And Utah has no systematic framework for evaluating whether any of it works, whether any of it is safe or whether any of it respects our students' privacy and our parents' rights. That is what this bill addresses. Some of my colleagues believe we should prohibit AI in our schools entirely. I understand that instinct, but Prohi prohibition does not pro protect students. It simply pushes the technology underground into personal devices with no oversight at all. Others believe we should let districts move forward on their own without state involvement. I understand that instinct too, but absent of framework, we are not enabling innovation. We are enabling a race to the bottom where the fastest vendor, not the safest one, captures our classrooms. This bill chooses a third path, not prohibition, not abdication guardrails. SB 3 22 creates Utah's education, educational technology, regulatory sandbox, a voluntary time-limited testing framework where AI tools can be evaluated in real classrooms under real conditions with real oversight. Before any decision about statewide use is ever made, think of it as a clinical trial for educational technology. We do not approve drugs for widespread use based on a vendor's promise. But that they work. We test them, we measure them, we evaluate them against rigorous independent standards. Then we decide we should expect the same discipline from technology being used with our children. Here's how it works. Before any AI system enters a pilot classroom, the vendor must complete red teaming adversarial testing where experts attempt to break the system the same way a curious teenager might. They test for harmful content. They test for emotional manipulation, they test for data extraction. They document what they found and what they fixed. No documentation. No approval. And for multi-year pilots, this is not a one-time certification. Updated testing is required annually. Safety failures must be reported to the state. Within 48 hours of discovery, the monitoring continues for the life of the pilot. No educator may use an AI system with students until they have completed required training at no cost in formats accessible to educators at every level of technical expertise. The first substitute made this explicit training first classroom. Second, parents receive written notice before AI interaction with their child. They can opt out at any time without academic penalty. Their child's education does not depend on their agreement to any of this. No ays, no AI system under this bill can assign a final grade, make a placement decision, or override a teacher's professional judgment. A licensed educator must review and approve before any AI output becomes part of a student's official record. The teacher is always the final authority. I wanted to say something about what this bill does for students specifically because it goes beyond what most AI legislation does anywhere in the country. The bill establishes what I would call a student due process and dignity framework. Students must be told that when they are talking to an AI rather than a human, they can ask out how any AI recommendation about them was produced. They can request that a human educator review any AI generated decisions before it takes place. They can contest those decision and they cannot be pen penalized for declining to use AI tools at all. AI systems under this bill are explicitly prohibited from simulating romantic or personal relationships with students from creating emotional dependency, from discouraging a student from seeking help from a parent, a teacher, or a counselor. And if a student anywhere in the pilot program expresses thoughts of self harm or suicide, the system must immediately con connect them to safe ut to the National Suicide Prevention Lifeline and to school counseling resources. The AI cannot act as a therapist. That role belongs to human beings. Perhaps the most important structural framework in this bill is the one that protects this chamber. No. A AI system can be adopted statewide in any school across any district without explicit authorization from this legislature, not from an agency, not from a board, not from a vendor. The state board of Education reviews the evidence from the completed pilots and makes recommendation to us. We decide the board cannot mandate adoption. It cannot condition funding on adoption. It cannot require specific vendors. The legislature retains final authority. That is not an accident. That is design. We are not going to take a vendor's word for it that their product works. Independent evaluators selecting selected competitively with no financial ties to vendors will measure actual student learning against the same rigorous standards. We apply to any other statewide educational initiative before and after. Assessment. Comparison groups. Disaggregated analysis for every protected student population, affect size calculations. That evidence comes to this legislature before we are asked to authorize anything at scale. One provision in this bill deserves direct mention. When AI systems support instructions in history, civics, or social studies, this bill requires them PRI to prioritize historically accurate period, relevant primary sources and to avoid what historians called presentism judging people of the of the past by standards they could not have known without grounding in the evidence of their times. This is not about prescribing conclusions, it's about methodological integrity. It is the same standard any serious historian applies. It is the standard we should expect from AI systems teaching our children about who we are and where we came from. No new appropriations, no new agency. The program runs through existing offices. Vendor vendors bear their own compliance cost and the leg entire framework sunsets automatically on July 1st, 2031, unless it's legislature votes. To extend it. Mr. President Prote, I have spent my career at the intersection of technology and data. I helped build some of the earliest neural network applications in this country in the 1990s, and I know what this technology can do. I also represent families in Morgan Summit and Weaver Counties who have been asking me all session to make sure that whatever happens in their children's classroom, somebody with authority is paying attention. They are not anti-technology. They are parents. They are asking exactly the right questions. This bill is my answer to both of those. If we do not build this structure now, it will build, be built later after something goes wrong in reaction to harm. That did not have to happen. I would rather build it right on our terms. Before that day comes. This bill allows Utah to lead deliberately, transparently, and constitutionally. I ask for your support on SB 3 22.

I move that SB 3 22 under suspension of the rules pass.

Vote. Senator Vickers Baldy Aye. Lewin Brammer Bus Colmore Escamia Fillmore Aye. Grover Aye. Harper Aye. Hankins Aye. Ipson Aye. Johnson Aye. Quan Aye. McKay Mikel Milner Musselman Owens pitcher plum Rebe Aye. Sandal Stevenson Stratton Oh two Bickers two 90 Weer aye. Wilson Winterton a Lewin

Thank you. Uh, Mr. President Prote, motion to encircle third substitute Senate bill two 90. The motion is to circle third substitute Senate Bill two 90, victim and witness privacy amendments.