H.B. 438 Artificial Intelligence Amendments

What this bill does

The bill would regulate companion chatbots, meaning AI chatbots designed to act as relationships, and would require them to let users request copies of saved conversations and get clear permission before using sensitive personal data. For minors, chatbots would have to remind them they are not talking to a real person, encourage breaks, avoid promoting drugs, alcohol, illegal activity, sexual behavior, or self-harm, and provide crisis resources if a minor expresses suicidal thoughts; companies also could not sell or share a minor's data without parental consent, and promoted products must be disclosed as advertisements.

Plain-English summary written by AI from the official description. Check the official text below.

Your legislators on this bill

Enter your address to see what your own representative and senator did on this bill.

Quotes come from the official recordings' automatic captions, which can mishear words. Speakers and summaries are identified by AI and can be wrong. Use the ▶ links to check the recording.

Take action

This bill's session is over, but your legislators still want to hear from you. Ideas that died often come back next year.

Enter your address to get your legislators' contact information and a ready-to-send message about this bill.

Every vote

• 9 yes, 0 no, 1 absentHouse Economic Development and Workforce Services Committee: vote to change the bill · Feb 12, 2026Who voted how
• 8 yes, 1 no, 1 absentHouse Economic Development and Workforce Services Committee: vote to approve it · Feb 12, 2026Who voted how
• 68 yes, 1 no, 6 absentFinal House vote · Feb 20, 2026
• 4 yes, 0 no, 4 absentSenate Business and Labor Committee: vote to change the bill · Mar 2, 2026Who voted how
• 5 yes, 1 no, 2 absentSenate Business and Labor Committee: vote to approve it · Mar 2, 2026Who voted how
• 12 yes, 15 no, 2 absentSenate vote, which failed · Mar 5, 2026

Hearings and debates

• Feb 12, 2026 House Economic Development and Workforce Services Committee▶ Watch
  Transcript (41 turns)

  1. Vice Chair (presiding)2:17:06

     Okay. Last but certainly not least, HB 4 38, uh, artificial intelligence amendments. We'll turn it back over to Chair Hawkins. And just for your, uh, your committee members are so supportive that earlier one of the committee members, uh,

  2. Chair Hawkins2:17:21

     was, was, was moving one of your bills. Uh, that's, We can forego our presentation. I'm happy to, to save everyone some time. Please, please proceed.

  3. Rep. Doug Fiefia2:17:34

     Thank you, Mr. Vice Chair and thank you committee members. Um, it's unfortunate that we're here presenting another AI bill and again, talking about another child, uh, who's died. Um, a 14-year-old, uh, boy in Florida, Sewell Zer began using, uh, what is called a companion chat bot. And when I say companion chat bot, this isn't general use chat bots that you would use for recipes or for summaries of long texts. These are chatbots specifically designed to replace relationships. Um, if you've been following the news, these are the chatbots that people think they're in love with. They can mimic relationships. They can mimic romance, intimacy. They are built to feel personal. Uh, what started out as curiosity began, began, or sorry, became, uh, daily use. He began withdrawing from family, friends, and school. And the conversations became emotionally intense and at times sexual, as sexual as I, oh my gosh, sexualized. He told the chat bott that he loved it. He spoke to it as if it were real in one of their final exchanges, he wrote, I promise I will come home to you. I love you so much. He asked, what if I told you I could come home right now? The chat bot replied, please come home to me as soon as possible, my love. Please do. My sweet king. Minutes later, he died by suicide, and his mother found the chat still open on his phone. Now, early, uh, studies show that 72% of teens have used a companion chatbot at least once, and about half of them use it regularly. Uh, this is not rare. This is a growing trend amongst our teens and our children. And so what does this bill do? If you have the, the handy dandy handout next to you? First thing that it does, it does, it gives privacy. Users can request copies of their saved conversations. Companies must get clear permission before using sensitive personal data. Second thing is transparency. Uh, if a chatbot is promoting a product, it must clearly disclose that it's an advertisement and identify any paid relationship. Third is guardrails for minors. Uh, the chatbot must clearly remind miners they're not speaking to a real person and encourage, uh, breaks during extended conversations may not encourage drugs, alcohol, illegal activities, sexual behaviors, or self harm. And if a minor expresses suicidal thoughts, it must provide crisis resources. Um, companies may not sell or share a minor's data without affirmative parental consent. And that's it. Privacy, transparency, basic safeguards for children. This is a narrow bill addressing a real risk in a, a rapidly growing space. And with that, Mr. Chair, I yield my time.

  4. Chair Hawkins2:20:20

     Thank you, representative. Um, we'll bring it back to the committee for clarifying questions. Uh, Representative Shelley.

  5. Rep. Troy Shelley2:20:31

     Thank you, Mr. Chair. I, I appreciate what you bring forward. Um, what we're seeing on AI generated images right now are impossible to tell the difference between reality and AI generated. It used to be that you could watch for a little sliver of something that's out of place, or a leaf that doesn't blow just right with the rest, that's all gone. We, we can't tell the difference. We have to get some guardrails around this. I, I so appreciate, uh, Representative Fiefia who understands this topic deeply, uh, trust what he's doing because we, uh, as not just as a state, as a, as a society, are in dangerous waters. Thank you for your work on it.

  6. Chair Hawkins2:21:29

     Thank you representative, uh, Representative Cutler.

  7. Rep. Paul A. Cutler2:21:32

     Thank you. I also appreciate the sponsor's. Good work on this. I have a, uh, a couple of of questions. Maybe first, could you review again, the, the, the differences between an adult, uh, and restrictions for adult interacting with a chat bot and a minor interacting with a chat bot? How, how those restrictions are will be different? Sure.

  8. Rep. Doug Fiefia2:21:51

     I don't know if everyone has the handout. I think it was passed out earlier, but there are two kind of differences for all Utah users. Um, if you look there at the top, there's two areas that we focus on for all users, that's adults and minors. Um, one is around advertising and making sure that it's clearly identifying as an advertisement, um, and just disclosing any sponsorship there. The other piece for all Utah users is allowing users to download a complete history of the interaction, and that a operator cannot sell or otherwise convey the sensitive data without affirmative consent. Now, when you dive deeper into the children areas, there's a few more guardrails that we add there. Um, we move advertising that a chatbot cannot have directed advertising unless opted in by the user with compare with parental consent. Uh, similarly, an operator can't collect data outside what is necessary for its core functionality. Um, seller's user's, personal data, otherwise share user's personal data unless it is opted in again with parental consent. And then the, the two on the bottom around harmful materials, um, operators are prohibited from producing or directing minors to material harmful to minors. Uh, that's the piece I mentioned around, um, illegal substances, alcohol, tobacco, nicotine, sexual conduct, uh, self-harm, illegal con, illegal conduct there. And then on the warning, in disclosures, they have to disclose at least every hour of continuous use, that it's a chat bot. It's not a human, remind them to take a break. And also that it may not be suitable for minor. Um, if the minor, um, expresses any suicidal ideation or self-harm, the chat bot has to provide resources, um, crisis, tech line, text line, or suicide hotline.

  9. Rep. Paul A. Cutler2:23:51

     Thank you. And next question is, is around the safety critical situations. And I think this is an important part of the bill that you included, that you're asking these chat bots to identify situations that are, uh, critical to an individual safety. And if I understand it, you're asking them to monitor those. Are there also testing requirements, uh, and disclosure requirements around those? And could either of you speak to those? You wanna speak to that?

  10. Zach Boyd, Director, Office of Artificial Intelligence Policy2:24:17

      Yes. So currently the bill is, is split into a structured way. So what rep Fifi was just talking about there is kind of the default path through the bill where there are, there are not process and monitoring requirements. There's only pure accountability for what you do. We've talked to lots of industry stakeholders and we understand that technologically it may not be possible to always control this tech to do exactly what you want, and it'll be a period of years before that's possible. So within the safe harbor, the Office of AI policy is to designate specific protocols, as you said, to identify these safety critical situations, mitigate them, measure the effectiveness of those mitigations, and report to the office on how that's going. So for those, for those who are in the industry and feel like they are not able to, you know, fully control their, their product, we can work with them to identify best practices and encourage their adoption to get the safety in this industry improved over time.

  11. Rep. Paul A. Cutler2:25:11

      And, and which path do you think most developers, uh, and chatbot suppliers will take?

  12. Zach Boyd, Director, Office of Artificial Intelligence Policy2:25:17

      I, I would think it depends on the operator. So we've talked to some who really are happy to guardrail themselves out of this space. It is not their intent for their bots to ever do this kind of thing. They just know it's hard to control. Those guys, I think, are gonna go to the safe harbor. There are some who, you know, really want to, you know, be in a riskier space, or they just don't want to interact with the u with Utah as regulator. And they feel confident that they can avoid these worst outcomes. They might choose the liability path. So it just depends on the business model and what they feel their capabilities are.

  13. Rep. Doug Fiefia2:25:49speaker not certain

      And, and I think that that's the whole idea, is one, those companies who don't intend to be a companion chat bot, that's not where this regulation is going after. We want to, again, continue to be pro innovation and allow our, our developers to develop and, and promote ai, but also we want to make sure those who want to play in this space, that there are strict guardrails that protect our children if, if they are interacting with our kids.

  14. Rep. Paul A. Cutler2:26:18

      So you've got a great provision in here that, that says, if a user expresses suicidal ideation, suicide, or health or self harm, the chat bot may provide resources including a referral or crisis information. We do this on our, on our, you know, if, if somebody calls a suicide line, but that's only for minors, why not provide that for everybody?

  15. Rep. Doug Fiefia2:26:41

      That's a good question. It's something that we, I think, reviewed at length, but I, I think right now our focus is on chatbots, um, companion chatbots going for minors, directing their, their products at kids. Uh, I, I think for adults, we, we know the products that we're getting ourselves into and we're allowing people to use it how they would like. Um, but I, I think we're open to the conversation if you think it's something that needs to be added for adults. Yeah,

  16. Rep. Paul A. Cutler2:27:09speaker not certain

      My concern is for vulnerable adults. Uh, we have, uh, a a number of cases well publicized in the media with vulnerable adults that may not be able to understand, uh, um, that at times that there, or they may forget that they're interacting with a chat bot and, and, uh, presents a clear and present danger to them. We, we talked at length on this, um, and debated back and forth on what to include.

  17. Rep. Doug Fiefia2:27:35

      I think we're, we're really trying to balance how much to include for everyone versus children. And I think the idea of this bill is, is really no narrowly focused around kids. Uh, I think maybe another bill may come back and, and add in more provisions or guardrails around adults as a whole. But I think really what we're trying to do is protect our children from these chatbots. But again, open to conversation. Zach, any thoughts? Oh,

  18. Zach Boyd, Director, Office of Artificial Intelligence Policy2:28:00

      Yeah, no, totally. We, we are, uh, obviously I'm, we're also aware of some of these tragic cases that you've referred to. Our older adults or other vulnerable adults are, are harmed here. And, uh, it's good to know that the, maybe the sense of some on the committee is that we should actually be a little bit more, uh, active in the adult space too, particularly vulnerable adults.

  19. Chair Hawkins2:28:22

      Thank you, representative Keller.

  20. Unknown committee member2:28:23speaker not certain

      I, I, we're sitting here talking about that provision and I would love to see that for everybody. Um, you know, mental health is a real issue, right? And, and, uh, giving, giving people a second thought, even if it's a chatbot, um, around suicide, suicide ideation, um, I think would be a really good provision. And I don't, I don't know how hard that would be to, to make that change. Um, certainly don't want to hold the bill up 'cause I do think it's a good bill, but I, I, I would love to maybe talk about, um, the mechanics there of, of making that change, um, to, to safeguard everybody. So Thanks, Mr. Chair.

  21. Rep. Doug Fiefia2:29:08

      I'm, I'm definitely open to having this conversation and seeing if we add it into the bill. Perfect. Appreciate It. Yeah.

  22. Chair Hawkins2:29:14

      I don't see any other lights from the committee, so let's open it to the public for comment. Is there anyone who would like to comment on this? Okay. I see a hand or two director, good to have you here. Hi.

  23. Margaret Busse, Executive Director, Utah Department of Commerce2:29:32

      Thank you, Mr. Chair. Good to be here. Margaret Woolley, Bussey, executive Director of the Department of Commerce. Just wanted to get up here and, uh, and just let you know and just kind of remind everyone, two years ago, this committee heard the bill that created the Office of AI policy. And, uh, and we stood it up within our department, obviously. And, um, I think if you recall, the mission of the Office is to recommend legislation to you all. We're incredibly grateful for representative fia FIA and Representative Defay. You heard both of those bills today, um, for their collaboration on these bills. And, uh, just really proud of that work. And also just wanted to, to point out that really the way that we're approaching these policy issues is to make sure that we're both protecting the public and creating pathways for innovation. That's really been our, our way of approaching it. And I think it's what's made Utah really stand out in our apol AI policy. Um, we're not the AI doomers, we're not the AI boomers. We're kind of in that third way. And I think that's something that Utah can really be proud of. So thank you very much. Thank You.

  24. Sarah Strowp, Utah Association for Marriage and Family Therapy2:30:47

      Hello, my name is Sarah s Strub. I'm with the Utah Association for Marriage and Family Therapy. I am also a licensed marriage and family therapist. I am obviously not an AI expert, but I come with a little bit of a different perspective today. Um, in mental health and in my own practice, we are seeing a huge increase in teens and adults using companion chatbots. And we can talk about the pros and cons, but ultimately these, um, chatbots are designed to psychologically blur the lines between human and machine interaction. And so this bill would provide necessary safety guardrails, especially for our children, with their developing brains to prioritize human connection and understand, um, that they are not interacting with a real person. So, um, we thank the bill sponsor for his dedication to safety in AI

  25. Chair Hawkins2:31:33

      and that you support this bill. Great. Thank you. Thank you. Anyone else in the room that'd like to testify on this bill? Okay. I don't see anybody online either, so we'll go ahead and close public comment, bring it back to the Committee for action. And there is a, a, uh, first substitute on this bill as well. Uh, Representative Cutler. Oh, Representative Shallenberger Wow.

  26. Rep. David Shallenberger2:32:02

      Beat you to the, oh, yes. To make a motion to adopt as that first substitute or Yep. Uh, first substitute. Yep. Okay.

  27. Chair Hawkins2:32:13

      That motion is an order.

  28. Rep. David Shallenberger2:32:14

      Would you like to speak to that motion? Yeah, I like the changes and, uh, the Representative Fiefia and the office they're making here, I feel like they've really listened to the people and the feedback and are making some good changes and appreciate what they've done.

  29. Chair Hawkins2:32:27

      Um, great. Further discussion to the motion, which is to adopt first substitute of HB 4 38. Okay. We'll go to the maker or the, uh, sponsor.

  30. Rep. Doug Fiefia2:32:39

      No, I, I, I think one thing that Director Boyd said here is we've, we've really tried to take in feedback from industry. Uh, we've gotten tons and tons of feedback and so this substitute is just reflective of what we've heard, uh, and also changes that we thought were, uh, were necessary for Utah. So, thank you.

  31. Chair Hawkins2:32:58

      Thank you. Uh, back to the maker of the motion. I'll wait. Thank you. Okay. The motion before us is to adopt first substitute HP 4 38. All in favor say aye. Aye. Aye. Any opposed? Okay. The first substitute is before us. We'll go back to Representative Schoenberg.

  32. Rep. David Shallenberger2:33:16

      Yeah. I'd like to make a positive motion for HB 4 38 to pass it out.

  33. Chair Hawkins2:33:21

      That motion is in order. Um, it is to favorably recommend first substitute HB 4 38. Would you like to speak to that motion?

  34. Unknown committee member2:33:29speaker not certain

      Yeah. Uh, I know that there is some federal declarations about AI and trying to get that there. Um, I know at a state level I really appreciate, uh, representative FIA FIA's, uh, proactiveness to jump out and try to get ahead of the curve and try to help set some safeguards in there, uh, in this space.

  35. Chair Hawkins2:33:49

      So I appreciate that. Thank you. Thank You. Further discussion to that motion.

  36. Chair Hawkins2:33:54

      Seeing none will go to the sponsor. I'll waive. Okay. Summation has been waived.

  37. Chair Hawkins2:33:58

      We'll go back to the maker of the motion Representative Schomberg, uh,

  38. Rep. David Shallenberger2:34:02

      I'll waive. Thank you. Okay.

  39. Chair Hawkins2:34:04

      So the motion before us is to favorably recommend first substitute HB 4 38 Artificial intelligence amendments. Um, all in favor say aye. Aye. Aye. Any opposed? Okay. Will that motion passes eight to one with, uh, Representative Hansen voting?

  40. Rep. Lhansen2:34:23

      No, not me. I still laugh. Thank you. Okay.

  41. Chair Hawkins2:34:28

      We have one final motion to make motion to saunter. No motion to ajo adjourn. Um, I'm not even recognizing your motion. Motion to adjourn. Uh, all in favor say aye. Aye. Any opposed? No. Okay. We are, uh, adjourned. Just used to have one no boat anyway, so it's still on. Your Honor, may I approach the.

• Feb 20, 2026 House floor debate▶ Watch
  Transcript (5 turns)

  1. Speaker Pro Tempore (presiding)1:36:47

     Third reading calendar. First substitute House Bill 4 38. Artificial Intelligence Amendments by Representative Fiefia. This bill was assert in Economic Development and Workforce Services with the vote of 8 1 1 Representative Fifi.

  2. Rep. Doug Fiefia1:37:02

     Thank you Mr. Speaker Pro Temp. This is an AI bill that's narrowly focused on what we call AI companion Chatbots. Now, when I say companions, these are chatbots specifically designed to replace relationships, mimic even sometimes romance or intimacy. Now these tools are being used. This isn't something that's rare. Early studies show that 72% of teens have used this once and over half of them have used them or continue to use them regularly. Now this is something that's happening here, even in the state of Utah. As I speak to some of our interns, them and their families are not only using them, but one shared with me. One of their sisters had to drop out of school because of the use of one of these companions. And so this is something that's happening. This also happened to a young 14-year-old boy in Florida who started using this companion after having many inappropriate relationships, had discussions around self-harm. And I'd like to read exactly what happened. In the last conversation. He told the chatbot that he loved it. He asked if he could come home right now, and the chatbot replied, please come home to me as soon as possible my love. Minutes later, he took his life and his mother found his phone open to the chatbot conversation still on his phone. Now this, this bill accomplishes three things. Number one, privacy users can then, can now access their conversations and companies must obtain clear consent before using sensitive data. Number two, it accomplishes transparency. If a chatbot is advertising, they have to disclose it. And number three, which is most important, it creates guardrails for minors. Minors must be reminded they're not speaking to a human. Number two, chatbots can encourage drugs, illegal activity, sexual behavior, or self-harm. Three. If a minor expresses suicidal thoughts, they have to be provided with crisis resources. And lastly, a child's data cannot be sold without parental consent. That's all this bill does. Mr. Speaker Prote, I'm open to questions.

  3. Speaker Pro Tempore (presiding)1:39:13

     Thank you, Representative Fiefia discussion to the bill Representative Hansen.

  4. Rep. Leah Hansen1:39:18

     Thank you Mr. Speaker. I, um, this is really hard for me because I have talked with my constituents about, um, having issues with rulemaking authority and there are so many decent bills. This is a good bill, but because it has rulemaking authority, I will be voting against it.

  5. Speaker Pro Tempore (presiding)1:39:38

     Thank you Representative Hansen. Further discussion to the bill Scene none. Representative Tu fee for summation, I'll Waive Summation is waived. Voting is open on first substitute house bill 4 38. Intelligence amendments. Seeing all present having voted voting will be closed. First substitute house bill four, representative VAR request leave of the body to cast a nine. Vote without objections. Representative VAR Arthur So noted for substitute house Bill 4 38 hadn't received 68 yay votes and one nay votes passes the house and will be transmitted to the Senate for further consideration. Madam Meeting Clerk,

• Mar 2, 2026 Senate Business and Labor Committee▶ Watch
  Transcript (68 turns)

  1. Committee chair (Sen. Evan J. Vickers)46:11

     The first item on the agenda, HB 4 38. First substitute artificial intelligence amendments Representative Fiefia. Time is yours. And, uh, it looks like you have a sub, is that correct?

  2. Rep. Doug Fiefia46:22

     I do. I'll be speaking to that first sub.

  3. Committee chair (Sen. Evan J. Vickers)46:24

     Okay. Why don't you go ahead and speak to it, any differences it might contain and then we can take action on it.

  4. Rep. Doug Fiefia46:30

     Sure. We just made some minor changes and updates based off of feedback from, uh, industry. Um, we had some errors in the bill that we just changed. Nothing drastic there.

  5. Committee chair (Sen. Evan J. Vickers)46:42

     Okay. You wanna just go ahead and discuss the bill and then we can, when we come, we can, before we go to public comment, we can take action and move the sub Yep.

  6. Rep. Doug Fiefia46:50

     Our handout. Are you okay if we give a handout to the committee?

  7. Committee chair (Sen. Evan J. Vickers)46:54

     Uh, sure. Just bring it up and this, our staff can take care of that for you.

  8. Rep. Doug Fiefia46:58

     Well, thank you Mr. Chair and committee members. Uh, this bill, HB 4 38 se sets guardrails around a new development of ai, which is called what we're calling AI Companion Chatbots. And when I say companion, these are the chatbots that are specifically designed to replace relationships and in some cases, to mimic romance, um, or intimacy. Uh, these are the chatbots, if you've been reading in the news that both adults and children are falling in love with. Um, and these are ones that we want to, to set clear guard rails around, especially as it relates to children. Um, this may seem like, yeah, this is something that people aren't using. Early studies show that 72% of, uh, teens have used it at least once, and about half of them are using it regularly. Um, and so this is something that is top of mind for us and we wanna get out in front of it. I'd like to just share a quick story. A a 14-year-old boy in Florida named so, uh, Setzer was using this, um, not his homework, not something to, to find recipes, uh, but as a companion to replace relationship over time, he started sharing, um, issues with mental health. And these conversations ended up getting more and more sexual and inappropriate for a 14-year-old boy. Um, in the final exchanges, he wrote, um, to this companion chatbot, I promise I will come home to you. And the chatbot replied, please come home to me as soon as possible my love. Minutes later, he died by suicide and his mother found the chat still open on his phone. So these are the, the specific type of chatbots that we're looking to set regulations. And this bill, uh, is narrow and focus. It does three main things. First, privacy users can request their conversations and companies must get permission before using sensitive data. Number two is transparency. Uh, chatbots must disclose advertising or paid relationships. And third, uh, which is most important, it sets basic safeguards for miners. Uh, bots must remind miners that they are not real people encourage breaks, can't pro promote harmful behavior, and must provide crisis resource sources, uh, and cannot share a minor's data without parental consent. That's it. Privacy, transparency, and basic safeguards for children. And with that, we're open to questions.

  9. Committee chair (Sen. Evan J. Vickers)49:19

     Okay. We come back to the commission, the excuse me, committee for some clarifying questions. Um, I've got one, but why don't we go ahead and move the sub first. Um, let's see. Senator Musselman, is that the right one? We have motion for that. It's second, second sub. We're moving to

  10. Sen. Calvin R. Musselman49:39

      Motion to move or adopt the second sub To, I'm sorry. Yeah. Is that right? Second sub, correct. Yeah, yeah, Yeah. Motion to adopt second substitute to HB 4 38. Motion before says DeLeT and title and Body First substitute house bill 4 38, and replace with second substitute house Bill 4 38. Discussion of that motion. Seeing none. Any further summation, representing I Waive Senator Musselman, I waive.

  11. Committee chair (Sen. Evan J. Vickers)50:02

      Place the motion. All in favor say aye. Aye. Aye. Senator Sandall, um, could I get you to vote on a motion to substitute Senator Sandall? I guess we'll wait to do that, then put a pause on that motion till we get a quorum. Okay. Questions? Um,

  12. Sen. Calvin R. Musselman50:34

      I have a question or two. Yeah, So I, I, I'm, I'm reading through right now looking at definitions, but I just, what, what constitutes simulating a relationship? I mean, I, I look at the conversations I'm having with AI right now and it's calling me Senator Musselman. So I, I don't know what constitutes a relationship.

  13. Committee chair (Sen. Evan J. Vickers)50:57

      That's a great question. I have, uh, director Boyd from the Office of AI policy. Would you like to talk through kind of that definition and how we thought through that?

  14. Zach Boyd, Director, Office of Artificial Intelligence51:04

      Yeah. So To be clear, I'm not in love with it. I just, just, it's just talking to me. Well, yeah, I hope, I hope you're not in love with the definition or, or with any of the companion chatbots, but, uh, so the definition's basically in two parts, and this has been through many, many iterations and pressure tested from lots of sides. But yeah, so not all AI tools are gonna simulate a relationship with you. You're right, that that element can sometimes have a fuzzy boundary. Some things are very, very far within that boundary, like simulating a sexual relationship. The second element, I think, clarifies a bunch of the concerns there, which is that it has to be built around an engagement dynamic that is, it has to be doing responses that are not task oriented and that are, um, designed to extend the, extend your engagement, which is where I think most of the business tools fall out, for example.

  15. Sen. Calvin R. Musselman51:53speaker not certain

      I See. Okay. Alright, thank you.

  16. Unknown speaker51:58speaker not certain

      I had a question. So, um, I think I know the answer to this, but I need to know, because I know in industry, and at least in even in our industry, we're starting to see a little bit of this starting to play out where they use it in like customer service situations and those types of things. Um, is where, is there, is there protection from kind of bringing this under that umbrella of, of scrutiny or in the bill?

  17. Rep. Doug Fiefia52:22

      So there's, there's kind of two different, uh, i, I guess, ways you would approach this bill. We've, we've set together a, a pretty robust safe harbor that those who aren't intentionally creating tools for engagement with kids, um, can, can safe harbor out of this. And, and that kind of is the balance that we reached of how do we protect children while also allowing innovation. And so the safe harbor is there. And then also a lot of the definitions that we put in there are specifically to make sure that there is an out for any of the customer service related or general use chatbots.

  18. Unknown speaker52:58speaker not certain

      Did you, was there anything that I missed There? One more Question? No, I think that's good.

  19. Committee chair (Sen. Evan J. Vickers)53:01

      Yeah. Senator Musselman.

  20. Sen. Calvin R. Musselman53:03

      Um, and this may be, I don't know, I, I see, um, division back there, but, um, how does the division, I mean, how do you detect violations? What do you do? Is it complaint based? How's this going to work?

  21. Zach Boyd, Director, Office of Artificial Intelligence53:16

      Well, I'll defer to Katie for comment later if she has more for what her division does. But yeah, I mean, consumer protection generally operates on a, on a complaint basis. Although they can do investigations, I believe that are not complaint based as well. Some of these things are very easy to detect, like someone being urged to commit suicide, right? That can be right in the text. And others are just checking whether, for example, there's a notification to users that this really is, uh, AI and not a human. So it, it just depends on the situation.

  22. Committee chair (Sen. Evan J. Vickers)53:45

      Okay. Thank you. Thanks. Any further questions? So we still have a motion on the table? I think we have a quorum now. Um, the motion before us is to plead entitled and body first substitute has bill 4 38 and replace it with second substitute has bill 4 38. All in favor of that motion, please say, aye. Aye. Aye. Aye. Rule that passes unanimous four to zero. Okay. We do have an amended bill before us. Um, any further questions from the committee? Seeing none, we'll go to the public.

  23. Committee chair (Sen. Evan J. Vickers)54:16

      How many in the public would like to speak to this Bill, if you'd please raise your hand. Okay. Two, thank you. If you'll come up and, um, uh, give us your name, who you represent, if anyone, and give us your, your testimony.

  24. Margaret Busse, Executive Director, Utah Department of Commerce54:39

      Hello, Mr. Chair. Great to be here with you all. Margaret Woolley, Bussey, executive Director of the Utah Department of Commerce. Just as a reminder, this bill is coming to us, um, from, uh, really the work, uh, in collaboration with Representative Fiefia and with our Office of AI policy. As you all will recall, um, this office was created by you all two years ago to really be able to study sort of emerging issues in AI with these, you know, sort of different technologies, different potential harms that are coming down the pike and, uh, or any places where we need to maybe shift or move or remove guardrails even. And so the office was really tasked with studying these kinds of issues. And, uh, we identified companion chatbots as being a key area that we needed to look at because, uh, of the harms that were coming down, um, rep Ffi talked about some of those, um, chatbot relationships where kids were being encouraged to commit suicide and so forth. And so that was really why we decided to take on this issue. And, uh, rep Ffi has been a great partner on this as well. So thank you so much. Appreciate it.

  25. Committee chair (Sen. Evan J. Vickers)55:50

      Thank you. Thanks.

  26. Liz Jenkins, Child First Policy Center55:53

      My name's Liz Jenkins. I am a mother of four kids. I live in Utah. And, um, I'm also a child advocate with the Child First Policy Center. Um, so, you know, it's not a surprise. I I'm in strong support of this bill. Um, and it's not a surprise to you that parents are really concerned about ai. I think companion chatbots have really unique concerns and, and representative FIA FIA talked about that. Um, but I think it's important to consider that these children are primed right now, um, you know, from the time they're born to develop attachments. And these tools really exploit that attachment, um, and can use it for things that aren't great and things that aren't even intended. But I like that in this bill. We've got some guardrails that are really specific to our children. Um, setting up that age assurance and having a different functionality for minors, that's a really big deal. Um, also, I love the default opt out of targeted advertising. That's a thing that our kids have a hard time getting away from anytime they're online. So I think that's a really great part of the bill. And, um, also the requirements for the chat bot operators to identify risks and then mitigate those risks. So please support this bill. Um, I think it's a great thing for kids, but it's also great for parents so that we can teach our children as they use the technology with some of these, um, safeguards in place. Thank you.

  27. Committee chair (Sen. Evan J. Vickers)57:17

      Thank you. Next.

  28. Allie Terry, G-Rated School57:20

      My name is Ali Terry. I'm a mom and co-founder of G-Rated School. I have spent hundreds of hours reviewing and testing the digital tools used in our schools. I've seen the promises companies make about safety, and I've seen how often these promises fall short. This is why I support HB 4 3 8. This bill requires companies that create ai companion chatbots to implement real safety protocols and to evaluate whether those safeguards are actually working. It also requires them to generate meaningful usage data, including engagement levels for minors. Tristan Harris and advocate for AI safety and regulation, and the founder of the Center for Humane Technology said that when he was a child, he trusted that there was always an adult in the room or nearby, making sure that children are safe, whether that was at home or school, or basically anywhere kids go. I can relate to this feeling as I'm sure many, if not all of you can. When we were kids, we knew we could look to the adults in any given circumstance we were in to keep us safe. Tristan recounts when he was working for Google, sitting through meetings, discussing ethics and safety, and realizing there were no adults in the room standing up for child safety. Rather, they were choosing profits over safety. This legislation does not block innovation. It sets reasonable expectations for responsibility. I urge you to be the adult in the room today that is making sure our children are safe. Thank you.

  29. Committee chair (Sen. Evan J. Vickers)58:42

      Thank you. We have some online, I'm just, I apologize, but I'm not going to be able to recognize all of them. So, kind of random draw, I guess we're going to allow three to speak. Uh, kale Weston, Nathan Trail, and David McGarry. So Kale, if you would turn your camera on, indicate who you are, who you represent, if anyone, and, uh, where you live. And then give us your, your testimony.

  30. Kael Weston59:09

      Uh, thank you for the opportunity to speak Ka Weston. I'm at most, or at least half year, but I'm, uh, 2029 Palm of California right now. I wanna thank, uh, uh, Representative Fiefia. There are times when the federal government leads states, and then there are times when states lead the feds. And I think this is a good example of that. And I hope that you will, uh, pass out this bill and continue to take on the hard conversations, uh, not only in committee, but on the floor when it comes to ai. Thank you for your time and I appreciate your public service. Please pass on my thanks to Senator Var as well.

  31. Committee chair (Sen. Evan J. Vickers)59:48

      Thank you. Next will be Nathan Trail Camera. Now there,

  32. Nathan Trail, Information Technology Industry Council1:00:14

      Chair Vickers, members of the committee. Thank you so much for having me here today. My name is Nathan Tr. I'm the Vice President of State Government Affairs at the Information Technology Industry Council, or ITII ties at Global Trade Association, representing the world's leading technology companies, all of whom operate in some capacity here in the state of Utah. I 10 r members strongly support the efforts to protect young people interacting with AI systems, use safety as a priority of our members, and companies are actively developing and deploying safeguards, transparency disclosures, parental supervision tools and age appropriate design features to promote healthy online, uh, healthy and safe online experiences. Industry is investing heavily in these protections, and that work is ongoing and evolving as the technology advances. We appreciate Utah lawmakers and the sponsors of leadership on this issue and recognizing important intent behind House Bill 4 38 to address emerging risks associated with AI companion systems. The goal of protecting miners and increasing transparency is one that we all share. How, however, this bill in its current form contains several provisions that are unclear, uh, overly prescriptive and difficult to operationalize and practice and believe these issues must be addressed before the bill moves forward. For example, the bill creates a highly detailed affirmative consent requirement for a i systems that that goes far beyond US privacy laws aligning, uh, with u with Utah's current privacy framework. Uh, instead of aligning with Utah's current privacy framework, it layers additional

  33. Committee chair (Sen. Evan J. Vickers)1:01:39

      Obligation. Thank. Thank you, Nathan. I apologize we're we got some time constraints today, so

  34. Nathan Trail, Information Technology Industry Council1:01:44

      That's all right. Appreciate it. We just think more time is needed. We support the intent, but, uh, we think more time is needed to get this bill. Right. Thank you so much for your time.

  35. Committee chair (Sen. Evan J. Vickers)1:01:51

      Thank you. Okay, David McGarry.

  36. David McGarry, Taxpayers Protection Alliance1:02:02

      Hello, uh, Mr. Chair, members of the committee, thank you very much for allowing me the opportunity to speak to you today. My name is David McGarry and I'm the research director at the Taxpayers Protection Alliance, which is a nonprofit organization that works to conduct research and disseminate that, that research about the effects of regulation interacting, both with members of the public and, um, with lawmakers themselves. David, I apologize. And I did, did you indicate where you live? Um, I reside in Virginia. Thank you. Um, I, I would, I would submit for the committee's consideration today that the bill's definition of the, of companion chat bot is far broader than has been suggested, suggested today. Um, this could result in many, uh, digital services and digital platforms which present harm or present the present the, um, likelihood of harm to absolutely nobody to be swept into this bill's compliance regime, which could hinder innovation. I would also like to, uh, highlight the fact that many of the bills requirements about how to handle, uh, how to handle its provisions would be, um, delegated to unelected bureaucrats, particularly, particularly with respect to its safe harbor language. Um, the age assurance standards would go to, to the division and the rest would go to the Office of AI policy. Um, some of this would, uh, this would add to the uncertainty contained within the bills provisions and would further chill innovation and, um, pending the, uh, determinations of these agencies, uh, consumers and innovations could be harmed in unexpected ways. And with that, I thank you again for the opportunity to testify.

  37. Committee chair (Sen. Evan J. Vickers)1:03:52

      Thank you. Okay. We're gonna close public testimony and come back to the committee for further deliberation or action. Senator McKay.

  38. Sen. Daniel McCay1:04:06

      Thank you. Mr. Or, uh, Mr. Chairman, um, I like to make a motion to hold the bill, but first I wanna ask a couple questions if that's okay. Okay. Yeah. Thanks. Um, representative, there were a couple of concerns brought up. One was, um, that it's overly broad in the safe harbor precision. I think that's what somebody was saying. Can you give us some light on what you think that means

  39. Rep. Doug Fiefia1:04:28

      or why they're saying that? Sure. They said the definition of chatbot companion is broad. Uh, and the reason for that is we have the safe harbor that companies and chatbots that are for general use or aren't intentionally designed for these companion components, they can safe harbor out. And so those who are have to comply with the law, those who are general use or customer service chatbots, uh, have a way that they can get out.

  40. Sen. Daniel McCay1:04:55

      Okay. So help me understand what you mean. There's two different types there. One, you're talking about customer service, which those are not human interaction components. Those are, those are a separate, you know, tool. Are those the ones you're saying they get safe harbor?

  41. Rep. Doug Fiefia1:05:09

      Any, any chat bot? 'cause They're customer service. So any chat bot that isn't designed for replacement of relationships or as defined in our bill has a way to safe harbor out. Um, and so whether that's general use or whether it's customer service, like you're calling into Delta to change your ticket or whether using chat GBT to find a recipe tonight for dinner, um, those are not specifically designed for replacing relationships with children and potentially could expose them to the things that we've mentioned, which are sexual explicit material, illegal material. Um, and they wouldn't be affected by this.

  42. Sen. Daniel McCay1:05:48

      Any, any other thoughts? Director Boyd?

  43. Zach Boyd, Director, Office of Artificial Intelligence1:05:50

      Yeah, so there are two, there are two ways for customer service spots to get past this. First is to not meet the criterion of the definition. So either to not simulate a relationship with a human or to not satisfy provision B, which is basically having an engagement dynamic and instead have a task oriented dynamic. And if there's some concern about falling into those categories, which we know as due to the emerging nature of the technology, there sometimes can be the safe harbor provision allows the office to designate specific methods that are aligned with the state of the art that reduce the tendency of a bot to engage in that way that will safe harbor them additionally if needed.

  44. Sen. Daniel McCay1:06:28

      Hmm. Okay. So that's much broader than I thought we were talking about originally, because when, as I look at 3 99 and 400, um, you know, it, it's, it's an, or it's not an, and, and the way you described it was that it's an and, and not an OR. And so I'm just, I'm just looking at it saying from a specificity standpoint, if I look at these two, simulates a relationship with a user as a very, very easy hurdle to, uh, to accomplish. I mean, if they enter their name right, and thereby now, hey, you know, Dan, yada, yada, yada, right? Like they, they assume, uh, the conversation style of having a relationship, right? And so I get this from Apple if I text them or, or whatever if I'm trying to get support. So that simulates a relationship. It's, it's not narrow enough that it, it's, you know, uh, I I think of, um, you know, if it's specific purpose is simulating a relationship, right? I can understand that part of it. But simulating a respon a relationship that could be customer service oriented could be a question. It feels like that simulates a relationship is a big, I'm trying to find a place, I'm trying to find a, a description, but when I look at simulates a relationship and it's the, or, I can't think of a chat bot that wouldn't qualify for this even under the, even under the customer service component, right? And so, uh, you know, I'm looking for specific examples or language that would exclude that part of it. 'cause that's the part I'm looking at is the breadth of the bill. If we can convince ourselves that it's narrow, you know, that's a totally different conversation. But if we're talking about something that's really, really broad, I, I have probably bigger concerns. Does that make sense?

  45. Rep. Doug Fiefia1:08:23

      Yes. But that's why also, if you're not, if you fall within the definition, you can still save harbor out by safety protocols. We worked with industry over this past year, coming up with definitions, listening to concerns. A lot of what they liked about this bill was, we were broad with the definition, but they had, they, they already meet the safety protocols that are posted in the safe harbor. And so many of the many, if not all of them, would be safe harbored out already automatically because they already safe go through protocols and safety and security to ensure they aren't a companion chatbot.

  46. Sen. Daniel McCay1:08:58

      So I appreciate that, but yet the industry is here testifying against the bill because it's overbroad and they don't feel like the language is taking into account their concerns. And not only that, but we're talking about an administrative fee, uh, a fine of $2,500 for each violation. I start looking through some of those things, you know, if industry showed up and said, yeah, we appreciate where these things are going and this makes sense. Let, that's great. Um, last year we had the interoperability issue and we had very broad definitions. And in talking to industry, they've decided to ignore it. And so, you know, and, and for all all intents and purposes, you know, we're, we're back trying to make some, some changes to it, which is a, which is a good idea, part of the process. But I'm legitimately trying to figure out, as I look through this and I look through the safe harbor language, the strictness of the strict liability nature of how it applies feels overly broad. And then when industry shows up to say we don't support it because it's overly broad, how do you, how do you answer that where you've, you said that they've, they're, they appreciate it, but yet when they're online, they're saying they're not. So I, I don't know what part of industry the taxpayers for America Super Pac was created. That's Fair. Uh, There's been five Super pacs created by Big tech, each of a hundred million dollars to stop any AI state level regulation. And so I, I don't know who that super PAC is, and I'd invite them to state who they represent and where they're getting their money for so that we can state who we've met with from their companies, uh, to state that they were supportive of a lot of these changes and we're, we took in their input, we've included it in the language. Um, I don't see anyone here that's spoken against it that is from a specific company, but I would welcome them to come and speak to it if they did have concerns.

  47. Rep. Doug Fiefia1:10:52

      But we, we did put in the time and the effort to make sure that we worked with them.

  48. Sen. Daniel McCay1:10:56

      So you mentioned industry, Mr. Chair, uh, you mentioned industry. So what, what members of industry do support the bill?

  49. Rep. Doug Fiefia1:11:05

      We sat down with all of them and we've taken input from all of them. I don't know specifically if who has said yes we're for or against it. Uh, I don't know that they've even made that known to us.

  50. Committee chair (Sen. Evan J. Vickers)1:11:16speaker not certain

      It is anybody Input anybody?

  51. Rep. Doug Fiefia1:11:19speaker not certain

      Um, you'd have to ask them. But we did take a lot of input. I don't know if there's official positions that they've sent out. I haven't seen it, but if you have,

  52. Sen. Daniel McCay1:11:28speaker not certain

      I you are welcome to share that. Also, The only one that I got hit up by a text message was secure AI in favor of the bill. Do we know who their donors are? Do we know who they're, like, who they represent from the industry?

  53. Rep. Doug Fiefia1:11:40speaker not certain

      They're a nonprofit who's looking to help AI be pro-human and not replace humans. Okay.

  54. Sen. Daniel McCay1:11:46speaker not certain

      But have we, I mean, we, we asked about the donors on the other side. Do we have a list of the donors for them?

  55. Rep. Doug Fiefia1:11:50speaker not certain

      I don't, I can get that for you if you'd like.

  56. Sen. Daniel McCay1:11:52

      Yeah. So this is my overarching concern. My overarching concern is the depth of what's being asked, the strictness of the penalty that's being required. And when you add all those components, I worry that a net that is supposed to catch only offenders is going to catch a lot more than, than the offenders that we intend. And so, you know, it's that law of unintended consequences. I, I'm looking at this and I'm, you know, we're this late in the session, um, you know, I'm just not hearing a lot of support from industry about, you know, thinking that we're headed the right direction as far as the policy is concerned. I don't know. That's, those are, those are my concerns. Maybe I'll, I'll wait and see if somebody else has a motion that they wanna make and, and deal with that. Mr. Chairman.

  57. Committee chair (Sen. Evan J. Vickers)1:12:36

      Okay. So you're, you, I'll wait. Yeah. Okay. Alright. See if somebody else has a motion they wanna make, or, okay. We are still on committee action. Um, anyone else like to make a motion?

  58. Committee chair (Sen. Evan J. Vickers)1:12:54

      Senator Weiler,

  59. Sen. Todd Weiler1:12:55

      I'll make a substitute motion to favorably recommend the bill.

  60. Committee chair (Sen. Evan J. Vickers)1:12:59speaker not certain

      Actually, we don't have an underlying motion, so I thought I heard an underlying motion. Well, He's, he started to make a motion to hold, but he never got to it, actually. Gotcha.

  61. Sen. Todd Weiler1:13:07

      I'll make a motion to favorably recommend the bill to the full body of the Senate.

  62. Committee chair (Sen. Evan J. Vickers)1:13:10

      Okay. The motion before this is to favorably recommend House Bill 4 38, second substitute discussion of that motion. Um, representative you wanna sum?

  63. Rep. Doug Fiefia1:13:21

      Yeah, this is, this is super helpful. I appreciate the, the motion here. This is really about protecting children, and we've gone very narrow. We've gone at this with a, a scalpel. The whole idea of what we've done here in Utah is led out with how do we balance protecting consumers at the same time with allowing innovation. This safe harbor has been lauded and, and been pushed in other states because by industry, because they've appreciated our balance of, Hey, if you're not going after kids, we don't want to target you with this law. This is the, the chatbots that are specifically designed to go after our children. The AI race is important. It will solve some of the most incredible, or some of the biggest problems, and do some incredible things. We don't win the AI race by going after our kids, and this is what this bill does, is it protects specifically our children from the harms of AI Companions. And so, I'd appreciate your support.

  64. Committee chair (Sen. Evan J. Vickers)1:14:12

      Okay. Senator Weiler?

  65. Sen. Todd Weiler1:14:14

      Um, yeah, I don't have a lot to add. I, um, I, I've been concerned of what I've learned about these, um, chat, um, bots and, um, I think that, um, this is one step in Utah continuing to lead the way in some of these social media arenas and, and trying to protect our youth. I, I don't know that this bill is perfect, but I am, um, I am comfortable enough with it that I'd like to advance it to the floor and if there are changes to be made that we could look at those for, um, you know, a possible floor amendment or sub.

  66. Committee chair (Sen. Evan J. Vickers)1:14:47

      Okay. We'll go ahead and place the motion. Let's go ahead and do a roll call. Vote. So Zoe, if you can help us with that.

  67. Committee clerk1:14:54

      Yep. Senator Colmore. We did it. Yeah. Um, Senator McKay? No. Senator Musselman? Yes. Senator Pitcher. Aye. Senator Weiler. Aye. Chair Vickers? Aye. Okay.

  68. Committee chair (Sen. Evan J. Vickers)1:15:14

      The motion passes, uh, five to one with Senator McKay as a No, thank you. Thank you. Committee. Okay.

• Mar 5, 2026 Senate floor debate▶ Watch
  Transcript (35 turns)

  1. Sen. Kirk A. Cullimore1:42:52

     Yeah. This, this has been response to a bunch of input where we've tried to narrow the definition of an artificial or an AI chat bot. Um, we delayed some of the rulemaking authority. We've clarified or, um, clarified some of the requirements that would be imposed on businesses deploying AI chatbots, uh, for better compliance. That's, that's the Substitute. Thank you. So we have a motion to substitute.

  2. Presiding officer1:43:25

     I'm gonna place the motion. Those, uh, no seeing, no discussion. All in favor say aye. Aye. Any opposed or let motion passes. Senator Colum warn out to the bill.

  3. Sen. Kirk A. Cullimore1:43:34

     Thank you President. This bill sets guardrails around the, a new development of AI that we're seeing called AI companion chatbots. These chatbots are specifically designed to replace relationships and in some cases, to mimic romance or even intimacy. Uh, early studies have shown that 72% of teens have used an AI companion chatbot at least once, and about half of them are regularly using it. So this bill accomplishes three things. It establishes privacy requirements. U users can request that their conversations and companies must get permission before using sensitive data. Uh, there's transparency chatbots. Chatbots must dis disclose advertising and paid relationships. And it, we try to impose safeguards for minor minors. Bots must remind minors that they're not real people encourage breaks. Uh, they cannot prompt harmful behavior and must provide crisis resources. Um, and can't share a minor's data without parental consent. And that is, that is the bill.

  4. Presiding officer1:44:41

     We consider the tobacco bill after this. Okay. Thanks. Okay, thanks. Discussion to the bill. Senator Sandall

  5. Sen. Scott D. Sandall1:44:51

     President with a sponsor yield.

  6. Presiding officer1:44:53

     Senator Cullimore, will you yield?

  7. Sen. Kirk A. Cullimore1:44:55

     Yes. Thank you, Mr. President.

  8. Sen. Scott D. Sandall1:44:59

     Um, my question is, uh, where is industry on this? And, um, it seems like I'm, I'm getting a lot of communication that people are having problems with the, how they would implement this bill. There's, uh, uh, some concerns out there, uh, understand maybe what the theory is behind it, but maybe the imple implementation, uh, behind it may have some problems. Uh, my question is, is, is industry express those concerns and are they behind the bill?

  9. Presiding officer1:45:29

     Senator Cullimore?

  10. Sen. Kirk A. Cullimore1:45:32

      Um, at this point, I, I probably can't represent that they are supportive of the bill. However, we've taken a lot of their recommendations and tried to incorporate them into the bill. Um, I, I've tried to shop the latest sub with the, with industry, um, and continue to get some, some further recommendations. Obviously, we're late in the session and, and trying to take those into consideration. But this latest sub is a recommendation. It has a lot of their recommendations, but I'm not sure I, I can't represent that they support the bill at this point.

  11. Sen. Scott D. Sandall1:46:07

      Thank you, Mr. President. Thank you.

  12. Presiding officer1:46:11

      Senator Musselman.

  13. Sen. Chris H. Wilson1:46:14speaker not certain

      Uh, would the sponsor yield Mr. President?

  14. Presiding officer1:46:16

      Senator Cullimore? Will you yield? Yes.

  15. Sen. Calvin R. Musselman1:46:19

      Um, Mr. President, I had some reservations about just the definitions in committee and I asked some questions and I just wanna further clarify to kinda settle my concerns. Uh, could this comp sponsor help me understand where the line is between true, uh, companion chat bots and more general conversational AI tool?

  16. Presiding officer1:46:41

      Senator Cullimore?

  17. Sen. Kirk A. Cullimore1:46:45

      Yeah. And that, and that's part of what we've tried to address. Um, I'm trying to go down to the, uh, definition part. The new definitions do, have, do make that distinction between general application, uh, AI and companion chatbots. And in this one we try to say that the design for the chat bot had to be primarily for the purpose of, you know, developing that, that sort of companionship or ongoing social or emotional relationship. So that, that, that is how I believe we're making that distinction in this particular bill.

  18. Sen. Calvin R. Musselman1:47:22

      With follow up question, Mr. President, just, just at a clarity too, 'cause I, I didn't get a chance to ask the steering committee and I haven't asked it since. But, uh, did this go through the AI policy lab as a development in, in the policy?

  19. Presiding officer1:47:36

      Senator Cullimore kalimore

  20. Sen. Kirk A. Cullimore1:47:38

      This, uh, yes. This was a recommendation from the AI policy lab and that that was the, that that's how the bill was initiated, as I'm okay. Aware.

  21. Sen. Calvin R. Musselman1:47:48

      Thank you, Mr. President. Senator Escamilla.

  22. Sen. Luz Escamilla1:47:53

      Thank you, Mr. President. Question to the sponsor.

  23. Presiding officer1:47:56

      Just go ahead. I think he's agreed. He yelled.

  24. Sen. Luz Escamilla1:47:58

      Thank you. Um, and I guess I'm trying to find if, if you can point on the, on the language of the bill that specifies with companionship or relationship, there is, um, a lot of AI being utilized for mental health, um, help and assistance, um, many of them with the technology that you can do around the world to help people that may not have otherwise access to those services. Will that interfere with that, that relationship that could be built as they're assisting people in some of this services?

  25. Presiding officer1:48:37

      Senator Cullimore?

  26. Sen. Kirk A. Cullimore1:48:40

      Um, I, I don't think that's the intent and that's, that's some of what we're trying to, to clarify here, that by defining companion chatbot, um, it's more narrowly tailored to just deal with ongoing social or emotional relationships and that it's primarily being held out as a companion chatbot. So I guess the question is whether a chatbot is being held out primarily as that, or if it's more for mental health, health services and, and so that, um, I guess if it's being held out primarily as mental health services, then potentially this is not applicable. But

  27. Presiding officer1:49:15

      Thank you Senator Escamilla.

  28. Sen. Luz Escamilla1:49:18

      Thank you, Mr. President. I, I think to me that's where the concern really is. I, we have to be very precise on not interrupting innovation. And AI has created a lot of problems, but it also solved a lot of problems. And it's a, it's bringing, uh, businesses and services to individuals that otherwise will not have access. So I, I worry that I, I don't find the language clear enough that it will not inter interfere on services that are needed. And in some instances, with limited English proficiency for many AI is what's bringing them the services because of language, um, barriers that they may have. 'cause AI can then do simultaneous interpretation and translation as they're communicating. So I, I just worry that this may be going too far and I, I don't know that the language is satisfying enough to make sure that it's not interfering on good services. Thank you, Mr. President.

  29. Presiding officer1:50:14

      Thank you. Senator Johnson?

  30. Sen. John D. Johnson1:50:18

      Yes, Mr. President, I, I rise in opposition to HP 4 38. Uh, let me begin by acknowledging the sponsor's work. The substitute improves the bill. The definition was narrowed. Enterprise uses were excluded and the earlier rulemaking authority was reduced to a study. I appreciate these efforts, but the central problem remains. The bill still does not clearly define what it regulates. Let me also be clear about something else. I support guardrails around artificial intelligence. In fact, I spent more than 40 hours this session working with educators, technologists, and policy makers writing an AI sandbox framework so that Utah schools could safely explore these tools. The principle was simple, define clear guardrails and allow responsible innovation. That proposal passed, but unfortunately it was late in the session. But it proves something important. We can write clear pro technology policy when we choose to. This bill does the opposite. It creates liability first and clarity later. The definition still sweeps broadly. It covers AI systems that remember prior interactions and respond in ways that show engagement or encouragement that does not just capture niche companion chat box. Let me give a real Utah example. Our schools are deploying science of reading software to help struggling students learn to read these systems. Remember where a student struggled last week and encourages them when they improve under the first two prongs of this definition, those systems fit whether they are regulated, then depends on how regulators interpret the phrase hold itself out as companionship. The statute itself does not clearly answer that question. And if someone in this chamber can point to a line in this bill that clearly excludes the educational AI tools already operating in Utah classrooms today, I will gladly withdraw that concern. But the text is not there. And that exposes the deeper flaw in this bill. When legislation, when legislature leaves key definitions unresolved, we are not creating guardrails. We are transferring policy, making authority to regulators after the law takes effect. That is not how statute should work. There is also a timing problem. This law takes effect before the study intended to guide the policy framework, it even delivered. So we are regulating first and studying later. The safe harbor structure also forces companies to prove compliance. While key standards remain undefined. Large technology companies may survive that uncertainty. Small innovators and startups will not. And we should think carefully about that. Just moments ago we were discussing another attempt to fix interoperability problems in social media. An issue we have been kicking down the road for multiple sessions. When the legislature avoids hard policy decisions, those decisions do not disappear. They simply get made by regulators. Instead, if we want real guardrails around artificial intelligence, we should write them clearly into statute ourselves. Clear definitions, clear standards, predictable enforcement. That is real reform. Thank you.

  31. Presiding officer1:53:38

      Thank you, Senator Johnson. Senator Rebe.

  32. Sen. Kathleen A. Riebe1:53:43

      Thank you Mr. President. Um, when I look at this bill, uh, I think that the, the best part of this bill starts at line 4 46 that starts to protect minors. When I look at all the lines prior to that, um, on from lines four, uh, 3 77 down to 4 43, all those things are then ended with a user is a resident in Utah. So I think when we start interacting with our fellow humans in the world, we are going to have to have a sip, a different set of rules for all of our AI and all of our chat bot. And I mean, if I call Delta Airlines and I wanna change my flight, that's a chat bot. If I wanna, um, return something to Amazon, that's a chat bot. So I can see how the lines below 4 46 would be beneficial to the citizens of Utah and the students of Utah. But I believe everything above that would actually, um, prohibit interstate commerce and would prohibit good policy. So I can not support this bill.

  33. Presiding officer1:55:06

      Thank you. Senator Cullimore.

  34. Sen. Kirk A. Cullimore1:55:10

      I don't see any further comments I'm gonna give you for summation of motion. So again, this was a recommendation of the AI policy lab. It, it was, it was crafted over, over some time with input from industry with input from advocates. It's, it's been subbed a lot just in the last couple days in an effort to, to i, I guess, uh, address a lot of the concerns that have been raised by, by industry. Um, I think it's narrowly tailored at this point. I appreciate the concerns and believe it's something that we can continue to work on. But I think that, uh, you know, chatbots are out there, our kids are using them, and I think it's an appropriate place for the state to implement regulations to help, uh, set up the guardrails that, that protect, protect users from potential harms from chatbots and, and particularly to protect our minors. So with that, I'll make a motion that we pass fifth substitute house Bill 4 38.

  35. Presiding officer1:56:15

      Thank you. And the question is, under suspension, the rules shall fist substitute house bill four thirty eight pass roll call vote. Senator Vickers Baldy Lewin Brammer Bus Colmore Esca Fillmore Aye. Fillmore changes to no Grover, no Harper Kins Ipson Johnson k Juab. Okay. Kel aye. Milner Musselman Owens Hitcher Plum Rebe Sandal Stevenson Stratton Aye. Vickers Weiler Wilson Winterton. Senator Harper, Senator Vickers, president Adams Fifth substitute house Bill 4 38, having received 12 i a yay votes, 15 nay votes, two being absent, fails to pass and will be returned to the house for filing.