Sign in or Join the community to continue
Event Replay: Responsible AI for Teens
Posted Oct 06, 2026 | Views 44
Share
Speakers
Chris Lehane
Chief Global Affairs Officer @ OpenAI
Chris Lehane is Chief Global Affairs Officer, Interim Chief Communications Officer, and a member of the executive team at OpenAI. Prior to joining OpenAI, Chris was Chief Strategy Officer/Operating Partner at Haun Ventures. From 2015 to 2021, Chris served on the executive management team of Airbnb and led the company’s policy and communications work. Before his time at Airbnb, Chris co-founded Fabiani & Lehane, a strategic consultancy that advised political, corporate, technology, entertainment, and professional sports clients. In the 1990s, he served in various positions in the administration of President Bill Clinton, including as Press Secretary to Vice President Al Gore and Special Assistant Counsel to President Clinton. Chris currently serves on the Board of Directors of Coinbase (NASDAQ: COIN), is a member of the Council on Foreign Relations, Trustee Emeritus of Amherst College and has served on numerous non-profit advisory boards. He holds a B.A. from Amherst College and a J.D. from Harvard Law School.
+ Read More
Kate Ruane
Director, Free Expression Project @ Center for Democracy & Technology
Kate Ruane is the Director of CDT’s Free Expression Project. An attorney with a strong background in legal research, Kate is committed to the freedom of speech and to bringing focus to the ways in which strong protections for free expression benefit communities of color, religious minorities, LGBTQ+ communities, and other oft-censored groups. Kate’s expertise is expansive and her work spans many issues including the intersection of civil rights and free speech protections, Section 230 of the Communications Decency Act, online privacy and surveillance, harassment, protecting children online, and disinformation.
+ Read More
Chuck Melley
SVP of Policy, Advocacy and Communities @ CodeAI
Chuck Melley leads policy, advocacy, and community work at CodeAI. He has more than 20 years of experience in education policy, including leadership roles at Pearson and ETS, service as a presidential appointee, and work on five presidential campaigns. His work has included advocacy across all 50 states and government relations in Europe, Asia, and Latin America.
+ Read More
Mathilde Cerioli
Chief Scientist @ everyoneAI
Mathilde Cerioli holds a Ph.D. in Cognitive Neuroscience and a Master's Degree in Psychology, with a strong focus on child development. As Chief Scientist at everyone.AI and iRAISE, Mathilde is committed to integrating responsible, safe, and ethical technology to support children's development. She is the lead author of several landmark publications on the impact of AI on children and adolescents, including the latest: "Mapping of GenAI Impacts on Child Development".
+ Read More
Arthur C. Evans Jr., PhD
Chief Executive Officer and Executive Vice President @ American Psychological Association
Clinical and community psychologist and health care innovator Arthur C. Evans Jr., PhD, is CEO of the American Psychological Association, the leading scientific and professional organization representing psychology in the United States. With more than 172,000 researchers, educators, clinicians, consultants, and students as members, APA promotes and disseminates psychological knowledge to benefit society and improve lives – a mission consistent with Evans' life work.
Before joining APA in March 2017, Evans spent 12 years as commissioner of Philadelphia's Department of Behavioral Health and Intellectual disAbility Services. In that post, he led the transformation of that agency's approach to serving a wide range of individuals with complex needs. The transformation of the $1.5 billion Philadelphia system has relied heavily on public health strategies that contribute to better community health.
+ Read More
SUMMARY
Chris Lehane moderated a conversation on how AI can support teens’ learning while protecting their safety and independence. Panelists discussed how AI can help students work through problems, why safeguards need to reflect teens’ development, and when systems should connect young people with trusted adults. They also explored teens’ privacy and access to information, alongside the training schools and educators need to support AI literacy. Speakers emphasized evaluating whether AI products are safe and deliver the benefits they promise.
+ Read More
CONTENT & TRANSCRIPT
00:00 Chris Lehane
This is a great panel. I think in some shape, way, or form, I've intersected with all of you in different lifetimes. I think, Mathilde and I do this circuit at this point, Paris, New York, DC, but thank you so much. And you've been such a great voice on particularly on the kid safety piece. Dr. Evans, my wife is an Evans. We were just trying to figure out where the connections... She's from Trinidad originally. We were trying to really figure out where we went back. But we first met in a much colder place at Davos. And they do these Davos meetings, they're like 10-minute increments.
00:39 Chris Lehane
I think we ended up meeting for about an hour because it was so interesting. But the thinking that you brought to the space is really great and we've really benefited from it. And then my friend from Pearson, we worked together when you're at Pearson and now you're at CodeAI and you've really been at the forefront of educational policy. And then Kate, who I just found out, had originally worked at CRS. For those of you who are Hill staffers, I joked that I thought CRS was the original AI before there was actually AI. Like you go over there, you ask for something and you get this report back two hours later by someone who was really, really smart.
01:17 Chris Lehane
So thank you for being with us and all the work that you do, particularly on the freedom of expression and now Center for Justice and Technology. I'll just say a couple words to kick this off and then really want to hear from the experts here. So oftentimes when we get into conversations about children and AI, there is a natural tendency at some level to almost fight what I would sort of call the last war, sort of the social media piece. That's understandable. I think humans just reference what their most recent experiences have been and are, and there's a lot of important lessons to take on.
02:00 Chris Lehane
I know, Mathilde, you spent a lot of time on this subject. I do think it's really important when we start to talk about these issues in the context of AI that we do recognize AI is a different kind of technology. It certainly can be used in social media, but it is something much more akin to a general purpose technology, something like electricity or the combustion engine, fire, the wheel. And because of that, I think it's even more important that we get this right. I think we obviously got some things right. A lot of things we did not get right in social media because this is so much bigger at least in my view, even more transformative than social media.
02:38 Chris Lehane
The imperative of getting this right at the front while the clay is still soft, I just think is incredibly paramount. And everyone up here on the stage to my right have been incredible folks thinking about this. We introduced a product, I guess, about two months ago now, Eric, right before the school year began, called ChatGPT. At the core of the idea behind this was on our platform you can treat adults like adults, but you should really treat kids like kids and that there's a real distinction. And so very quickly, Chat for Teens, we do everything from an age assurance piece to making sure that there's a model designed for kids who are under 18.
03:24 Chris Lehane
It has all sorts of aspects to it, including notification that you're actually dealing with artificial intelligence, parental tools so parents can be able to monitor and intersect with it. But it also has this piece that is to me really incredibly interesting, we'll get into this here just more broadly, the concept, which is that it has the study mode, and it won't give you the answer. It will inquire and push back so that you actually have to work your way through the answer as a student or as a young person. And I do think that that speaks to both sides of this coin that we really have to get right, which is both the safety and responsibility.
04:07 Chris Lehane
These people do want to have confidence. Parents want to certainly have confidence in these tools, but also to make sure that we are thinking about the opportunity of all of this. You know, as I said, I'm a history major, and one of the things that really goes through my head when I think about the history of all of these things. So bear with me for two seconds. But there's the printing press, to the extent that Gutenberg created the printing press late 1490s. And two parts of the world sort of approached it in really different ways. Europe, because in many ways Europe was balkanized because there were different kingdoms, and the Catholic Church had created a baseline of literacy.
04:51 Chris Lehane
Europe really leaned into teaching people how to read, what was appropriate, what wasn't appropriate. There were a bunch of fits and starts, but ultimately really leaned into the literacy piece of that. And as a result, you went from that Renaissance, Enlightenment, Age of Discovery, and then you compare that at the time to China, which was under a dynasty, and there was a concern that the written word literacy would scale the ability of folks to have liberty to think, to be individuals, to create really what was at the center of the Renaissance, and they really cracked down on the ability of folks to be able to become literate.
05:30 Chris Lehane
Those two places went in very different directions for 500 years, and I do think there are some lessons here that we can take certainly from that time period. History rhymes, does not repeat, as is often said, but can we get the safety piece right? And if we get the safety piece right, then we actually unlock the opportunity here for kids to be able to learn, to think, to build, to create, and to produce. And so with that opening, Chuck, I'd like to sort of come over to you as someone... I think you and I go back to the Commerce Department.
06:04 Chris Lehane
You have much better hair. Your hair has survived much better. But, you're at Pearson for many years, which does educational textbooks and assets, and then obviously at CodeAI now. How do you... Where do you think that the highest opportunity is in terms of using this technology for that opportunity? Where do you see kids using this the right way?
06:24 Chuck Melley
Right. Definitely. Thanks so much. Great to see you all. Thank you for including me here. I really appreciate having the opportunity to speak with you all. Yeah. I mean, as you mentioned, we are at a seminal moment here in society and AI brings such great opportunity for many people. When it comes to learning, especially for children, it can really help them deepen their thinking as long as it doesn't give them the answer. And to the point you were saying earlier, it has the opportunity, if used correctly, to really help them think through challenges, think through computational issues, and think through how to get the best out of a data set and things like that.
07:05 Chuck Melley
I mean, it really, really offers them a great deal of opportunity to learn more and to dig deeper into things. That being said, they have to know how to do so. Yeah. So it's incredibly important to equip learners with the ability to learn, to become fluent, to become digital literate, or AI literate, so that they have the ability to flourish and succeed in that area. And also, when they become literate and fluent, they know what's wrong and what's right as far as cheating is concerned. I think that's one of the things that we're all afraid of as far as, so say I wrote my paper and I turned it in.
07:45 Chuck Melley
Well once the kids really know how to use these tools and applications, then they will know what is good and what is not good. And they themselves can tell the difference and it's a blocker between them and turning it in to the teachers.
08:00 Chris Lehane
Can I do a quick follow up with you on that, which is as you think about that cheating question, because I get that all the time. I'm sure everyone on this panel gets that all the time. I'm sure parents here [inaudible]. If you're thinking about that, how much of the cheating can be addressed through the tools themselves versus or maybe in balance with the ethos I think that you're getting at. I'm curious how you think about that.
08:30 Chuck Melley
Yeah, it's both. Well, it's all of the above quite frankly. Not only is it the tools, but also it's the ethos, it's the teachers, it's the parents, it's the learners themselves changing and it goes back to education, letting them know that they're not going to really learn this if they don't do it the proper way, approach it the proper way. And I think we see some signs of us getting there, but it's going to take a little bit longer I think for parents to demonstrate it, teachers to demonstrate it, and we appreciate what a lot of the organizations are doing as far as helping with that from a technological standpoint.
09:01 Chris Lehane
Dr. Evans, I'll come to you on this. At the beginning I had talked a little bit about social media and you spent a lot of time and probably as much as anyone in the American Psychiatry Association, as much as anyone on this, what are the right lessons, well let me rephrase that, what are the lessons that we should really take that we should be applying to AI from social media and how do you think about that? Like how do we fight the next war based on the last war?
09:30 Arthur C. Evans Jr.
Yeah, there are a lot of important lessons. I think about this from a perspective of protect, strengthen, and advance. One of the things we learned is that for the scientific community we needed to have acted much sooner and we should have acted on the evidence that we had. The reality is that if you look at the issues that youths are dealing with and that we're dealing with when it comes to these technologies, there's a continuum. There are parts of... there are things that we know we've done studies, multiple studies over decades, there are things that we know a little bit about, and then there are things that we don't know anything about.
10:10 Arthur C. Evans Jr.
And good scientists don't talk about things that they don't know anything about, right? Or they don't talk about or go beyond the data. The problem with that is those are still questions that we have to answer. And I think one of the things that the scientific community is much more poised to do now is to act on what we do know, and for those areas where we don't have the data, we need to develop the data, we need to work with the industry to get that data, and to come up with those answers. But we can't sit back and wait because AI is the largest experiment on the human brain and behavior in the history of humanity.
10:47 Arthur C. Evans Jr.
There's nothing that compares to that. And so we can't wait until we have answers before we act. We have to act on the data that we have. Second thing is in terms of strengthening, if you look at even your own mission, a lot of the mission statements they talk about, AI should be really improving humanity. And I think we have set the bar too low, we've set the bar at a do no harm level. It really should be when youth are interacting with these technologies, are they better off because they've interacted with these technologies. I think that's where the bar should be, and we need to think about what that means in terms of the daily use.
11:32 Arthur C. Evans Jr.
And then in terms of using the technology, one of the things I get concerned about is that we're so focused on fixing the, addressing the issues and concerns, that we are not looking at the opportunities like what you've talked about. If you look at mental health, for example, there are most kids who have a mental health condition in this country are not getting any care. And for, and I'm sure there are people in here who have youth that they've tried to access care, it's at least six months to a year before you can see someone in most places in the country.
12:10 Arthur C. Evans Jr.
Can this technology help us, not to replace humans, but to supplement and to do those things that the technology is actually good at, psychoeducation, those kinds of things. So I really think we have a much better handle on how we're thinking about the technology, how we should be thinking about the challenges, but also how are we thinking about the opportunities.
12:33 Chris Lehane
Okay. I'm going to come back here for a second. This will set up a question for, we have a lot of doctors, Dr. Cerioli, but just to come back to you. I found myself at a dinner on Friday night that I didn't really understand what I was walking into. It was a neighbor's, and one of them was a Harvard professor, but he had a bunch of other faculty members over, and it ended up 50 questions to me. But one of the questions that we really got into was on the kids' piece. And there was someone from the public health school there who really went deep on this whole concept of how a child's brain, and their point was look, kids' brains, I won't do this right.
13:15 Chris Lehane
You'll correct me, but some version of how your frontal lobes actually sort of evolve, particularly when you get to your late teens and early twenties, and sort of having to be sensitive to understanding that before that takes place. I don't know if you're comfortable sort of touching on that, but there is, there's actually a physical aspect to this, right? There is biology behind this.
13:35 Arthur C. Evans Jr.
Yeah, yeah. The best way to describe it is that for adolescents, developing youth, their reward system is very highly sensitive. They're very sensitive. That's why they pursue likes and those kind of things. At the same time, the prefrontal cortex where there's decision making and making judgment, critical decisions, that's least developed. That doesn't develop until you're about 25. So you have the perfect storm in these technologies that are built on the idea of keeping people engaged. So as one of my friends calls it, it's all gas and no brakes. So they're pursuing the engagement, they're pursuing the interactions, but they don't have the critical thinking skills. That's the thing that we have to be really concerned about and that's another important lesson for them.
14:23 Chris Lehane
Yeah. It's interesting you say that because Chuck goes back to the ethos is one part, but you have to build the systems in. Dr. Cerioli, you have done as much thinking on this really as anyone. I think a lot of your thinking has informed the EU and other governments around the world about what is, quote unquote, age appropriate. So I know you've done a ton of research on it, but could you speak to that a little bit?
14:47 Mathilde Cerioli
Yes. So I think, first, thank you for having me. The question about age appropriate and as you perfectly mentioned, so you really got the message from that professor, is why do we need to do things differently with kids? It's because their brains are developing, which means the experience they get at that age, shapes who they are, and they go through what we call those sensitive periods of development. All throughout the first 25 years of their lives. And so, the role AI plays during those periods can either be really to support, increase autonomy and get them to mastery, or they kind of learn around AI, but then if you remove AI, they don't know how to do it.
15:29 Mathilde Cerioli
So that's what we want to avoid, basically. And so that's really the approach we took for the mapping we had discussed back in June under President Macron. He asked to help from the scientific community actually to map what are the benefits and risk, and that's exactly to your point, which was we don't have all the evidence, but we know so much about what our children going through at which stage, and based on that, what are the design choices that seem like the right ones for adolescents specifically? As you mentioned, yes, they're more sensitive to reward, and on top of that, it's really an age that's very sensitive to social cues, because when you think of it, it's puberty.
16:12 Mathilde Cerioli
They need to kind of leave a bit of the parental nest and go out in the world, explore, and make those connections. So any social cues is especially attracting to them, and AI is anthropomorphic. It uses language. It really plays into that, so we really have to think, how do we design a system that will be anthropomorphic? And we anthropomorphize chairs, so we do that with everything, not just AI, but AI kind of can reinforce that, where we feel like, oh, I feel like the AI is thinking that, and then you can have the AI say, yeah, I actually feel like that about this. So how do we design through that, and how do we adapt what AI does and the behaviors to a child's age?
16:56 Chris Lehane
Yeah, the anthropomorphic question is really interesting. I had a piece out today, really trying to make a distinction. We don't believe that AI has divinity or spark of life in it, but we do recognize that it is. I mean, I did my little history lesson on general purpose. This is a different kind of general purpose technology and super transformative, and you can understand how people may approach it that way. I mean, there's a lot of folks in this room who deal with policy. You know, they have the benefit of being in the room with you today, and you've given a lot of policy advice around the world.
17:38 Chris Lehane
What's some advice that you would give? I have a, you know, we suspect, and we certainly are gonna advocate for it that we think of, you know, kid safety regulations at the national level. We've been pushing these at the state level, supporting activity in the EU. But for folks in this room who may be either from the advocacy perspective or actually working in congressional offices or on the Hill, like what would be your high level thoughts on how they should be thinking about regulations and policy here?
18:07 Mathilde Cerioli
I think we are at the stage where we need convergence on how do we operationalize high principle that I see emerging really across the globe of, the first one I think we're pretty clear with young users and AI is, AI should not pretend to be human. But what does it actually mean? Because what we're seeing, and models have really improved on that on the content perspective, which means if I ask an AI, do you want to be my friend? They're gonna say, I can't really be your friend because I'm an AI. But no matter what, you can come back to me and I will be there to answer all your questions.
18:43 Mathilde Cerioli
So then we have kind of a disconnect between what it says and actually how it interacts and how it behaves. And so those are some of the things we're actively working on the assessment piece at Everyone.AI, which is looking at 15 behaviors, what we call anthropomorphic cues. So anything that makes AI feel human, interactional, how it interacts to be validating, asking questions, and then the relational, and understanding at what age can we do what, and in what context, because you want a proactive model when you're learning, you want the model to be like, you failed that test, what happened? If a teen the day before said they wanted to self-harm, you don't want the model to say like, how did that go for you?
19:27 Mathilde Cerioli
So that's really how we need to think of it in context. And so what we do while the evidence builds at the policy level, is we ask researchers to tell us what is the right level of those behaviors? It's not a all or nothing, it's really a gradient. So on that scale, what feels appropriate in which context? Then we can use that same scoring system to actually build the research so that we have the evidence. So we use provisional threshold to help product teams like you, to help policymakers define what feels appropriate. And then when the evidence builds, we can move those gradients if we see maybe we were too conservative or not enough.
20:03 Mathilde Cerioli
But we have to have that proactive approach to safety. We can't just launch and then hope for the best. We really have to interact and to be there much earlier in the conversation. And I think that's really the lesson that I'm really glad all scientists have taken, I think, where they're starting to talk even if they don't have the evidence. And I feel when we have a gap, we should still ask the people who are the most informed about the subject to guide the safety on this topic.
20:33 Chris Lehane
Yes, that's fair. I mean, there's a great brain trust here for those in the room who are looking to develop and build policy. Kate, maybe coming to you, we're talking at the policy level, but policy ultimately can sometimes get translated to the breakfast table. I'm getting to ask you guys all the questions I get asked all the time, so I appreciate you guys serving this purpose. But I get asked a lot, okay, I'm a parent. I have kids who are maybe teenagers. I may know a little bit about this, but I know less about it than I even know about social media. If you're at that breakfast table or that dinner table and you're a parent and you're talking with your teenager, how would you approach that?
21:19 Kate Ruane
How would I approach talking to my hypothetical teenager?
21:21 Chris Lehane
Yeah, yeah, hypothetical teenager. No, but like a family approach, because we're talking at the highest, policy at the regulatory level is one thing. Policy within a family. and how they think about this.
21:30 Kate Ruane
Yeah, within a family I think it's much easier than at the higher level, right? Within a family, especially if the family is supportive and everyone is at the table and talking to each other already, we are in a better spot than what you necessarily are dealing with at the policy level where you're trying to deal with everything. But within your own family, you know what is a supportive environment for one kid versus another and one kid might need one thing and another kid might need another, and you can set rules on that basis. And for me, I think that the idea would be how can this support you at this time?
22:07 Kate Ruane
What is it that you need now in your life that this could be helpful for? Or maybe not. And being in conversation with your kid about what they are and aren't doing from a breakfast or dinner table place of supporting the rights of my own child, which sometimes I don't think policy really does. I don't think policy takes the position right now of supporting the rights of children, their right to privacy, their right to free expression, their right to access in education, their right to play. A lot of times we are seeing like a fear-based policy and I think that we also see similar fear-based policies at a family level too. And how we change that conversation to one of support and how we're using opportunities to support every child and support individual children is the question I would want to see us asking going forward.
23:04 Chris Lehane
Yeah, it's great. As you're well aware, we put in a bunch of parental tools to give parents... I think one of the things that has been a challenge certainly speaking from the OpenAI perspective has been something that you are really hitting on, which is there are kids who come from a variety of walks of life who end up really depending on the technology. They may not have the family support or the access to help. We probably ended up, not probably, we did, ended up being more on the blunt instrument side of trying to really limit what a child could get on that side because we really wanted to optimize for trying to keep it conventionally as safe as possible.
23:52 Chris Lehane
I do think, and Dr. Evans, you and I talked a little bit about this in Davos, but Kate, coming to you, how do we start to think about that over the long term? Because OpenAI has 1.2 plus billion people who use this on a regular basis, and people are using it in ways to help them deal with issues in their lives and obviously want to be super sensitive to the kid piece. Even that balance of privacy, that balance of that child feeling like they can get stuff that's appropriate, how should we be thinking about developing policies as we go forward on that?
24:26 Kate Ruane
I think this is one, this is obviously one of the questions-
24:28 Chris Lehane
It's a really hard question, by the way.
24:30 Kate Ruane
... questions of our time for a good reason, though, because all families are different and all children are different. At CDT, what we have done in the past, in the social media context, not to fight the last war right now, but I think this is translatable here, is we did research where we actually talked to teens and to their parents and said, what is it that you need? Because it is super important to have expert-informed opinions about how to design things, and I'm really interested and fascinated. But it's also really important to have the information from kids and parents themselves, where kids say, I want the opportunity to go to my parents when I'm having trouble, or I want the opportunity to go to a trusted adult, but I don't want the company to tell me when that happens.
25:14 Kate Ruane
And I don't want the company to be the one that's responsible for doing that. Instead, I want a transparent way to tell the company that I'm not happy with this, or a way to have the company go and tell my parents, or a way to facilitate that conversation, but not to have a one size fits all rule because all kids are different. And sometimes those tools can be used for abuse, right? Sometimes they can be used as tools of coercive control, or as tools to suppress a kid's ability to access the information that they want. So thinking that through, and I actually think that OpenAI has tried to do that to a certain degree, which I appreciate.
25:51 Kate Ruane
And I believe that kids have to consent to parents being able to see things like that, thinking things through that way, I think is really important. And then having data to show us what happens once that has been deployed, so we can iterate on what we learn.
26:07 Chris Lehane
So I'm going to take that really hard question, make it even harder for Dr. Evans, but he's personally up to the task. Going back to the first conversation you and I had, like how do you deal with a teen who may be in crisis? Again, a lot of policy folks in here, what should be the default mode? How should we be thinking about that? And I know you've given this a ton of thought.
26:30 Arthur C. Evans Jr.
Yeah, and I think it starts with the detection issue. I don't envy you all, it's very complicated to detect when a youth is in distress. You have a lot of dimensions. One is developmentally, what a 13-year-old says is different than a 17-year-old says and how you detect there. You have cultural differences, you said that a billion people are using your technology. Well, most of those people are outside the U.S. And even within the U.S. we have a lot of diversity, so that if you are talking about a youth in a rural frontier America versus a kid who's in a different part of the country.
27:11 Arthur C. Evans Jr.
And then you have this challenge of a context window that can be pretty large. It's fairly, I won't say fairly simple, but it's a lot simpler to detect if a youth says, you know, I'm thinking about hurting myself. It's pretty straightforward. But if the issue of distress emerges over a one-hour period, that's a lot harder to detect. So, you all have a lot of challenges there just in figuring out who are the youth that you have to be concerned about. The one thing that I would say is that the technology really should be a bridge to a human, a person who is trusted in that youth's life, and that that bridge and that connection should come before the youth is in crisis.
27:56 Arthur C. Evans Jr.
So, clearly, you know, if someone says, I want to hurt myself, it's a lot easier to say, okay, you need to connect. But I think the challenge is even as these issues start to emerge, I think that's the point at which you really need to make those kind of connections.
28:14 Chris Lehane
That's great. So, Mathilde, one of the things I think we often think about, and you've spoken to this a bit, you know, this should be, and all of you have, you know, evidence-based, but from a policymaker perspective, like this is still a relatively young technology and it's also accelerating. How do you think about, or how should policy folks think about the fact that, like, what we know today is likely to change, you know, and policy tends to be fairly static? I do think the EU has actually done a pretty good job of thinking about this in a little bit more of a flexible way.
28:46 Mathilde Cerioli
I think the first thing they've done is actually really position safety first. And then I think, especially when it comes to children, it's going to be always much easier to relax. As we go, companies are going to be much more willing to relax on their model rather than the opposite. So having, I think, the safety first approach, the ideal would be the beneficial first approach, that we're not just doing safety, but therefore really like showing that models are doing what they say they do. So I think we're seeing that the EU wants to move also in that direction of not just doing the baseline that is safety, but if you say your product is good for learning, where's the evidence?
29:28 Mathilde Cerioli
Where are the data? Which is what we do for many powerful products when it comes to medicine, for instance. We don't just launch a product and say, we'll see. It seems that there's no countereffect, and so it's going to be fine. I think EU and many places are moving in that direction of showing the proof that it's actually safe before launch. And I think, of course, there is a balance to find there between having to show everything that is safe, you will never be able to do that completely first. There's still like baseline requirements that we can meet. And I think there's also a way where in different places than with social media, where there's a lot of public interest.
30:19 Mathilde Cerioli
Social media was so new, like we didn't really know that what happened in the digital world actually transferred in the real world. We have mental health crisis for teenagers. So parents are looking at companies and what if the competition was who gets to be the safest? Who gets to be the product we pick? Because we know that's where we're comfortable putting our kids. So I think that's one of the difference as well is that with models, people are going to want to stay with the same model. Basically, they're pretty loyal. Whereas social media, you just go from one platform to another. So I think that can become actually the incentive, like be the best...
30:54 Chris Lehane
That does align incentives, right?
30:56 Mathilde Cerioli
Yeah.
30:57 Chris Lehane
That's a really interesting... Chuck, you spent your professional career dealing with schools and educators and the actual institutional side of this. And over your career, really, you came over the digital age and now the AI time period. What's your advice for how an institution or an educator based on your experience should be thinking about this from the adoption, adapting perspective?
31:22 Chuck Melley
It has been a while. But the way I look at it, it's a team sport. It's not an educator. It's not an administrator. It's the state, it's the districts, and it's the education community overall. As a matter of fact, one of the things, I just want to go back to how the OECD is really thinking about things moving forward, one of the first things I did at CodeAI is went to Brussels to look for the release of the OECD EU AI literacy framework. A lot of letters, but it's really important. Basically, that was the basis from which we thought, they thought, and everybody thought, the things that kids should know and learn.
32:07 Chuck Melley
It was created to be static, knowing that tools are going to change, the technology is going to be advanced. The skills and attributes necessary to succeed. A lot of those were things that made us human. It's like critical thinking and reasoning and things like that. I just wanted to point out that's a responsibility, I think of overall the educational system, but as far as what should ... right now, the way I look at it as far as the states and districts and the rest of us. As far as states are concerned, they should be doing that work right now as far as putting together their AI policies and their recommendations and their guidelines and things like that.
32:44 Chuck Melley
Standards, etc. Funding teacher professional development. I mean, because there was a study, I don't remember the exact numbers, but it's something like 83% of students are using AI in the classroom, but I think the school leaders said that only 12% of teachers are being trained on it. And so it's super important that states set the policies, fund the professional development, and then the district levels, they implement it with fidelity. It's like localized, you know, as far as the tools are concerned and the products and the methods of instruction while remaining true and faithful to the intent of the state overall.
33:21 Chuck Melley
And as far as the rest of us are concerned, it's really to, you know, back to what you're saying is it's really support, but also empower. Yeah. You know, provide these students with the tools necessary and the environment in which they can learn, you know, with as much flexibility and freedom as possible, but safety as well, to really move to that next level.
33:44 Chris Lehane
Yeah. You set Kate up really well there. Kate. Let me go just a little top line on this, which is one of the things we talk a lot at OpenAI about the importance of making sure that the technology is small-d democratized, meaning as many people as possible can access it as opposed to having it concentrated. One of the, at least, items that has caused me some real concerns, particularly in the U.S., this hasn't been the case in other parts of the world, but you've seen some school districts in the U.S. take a fairly draconian approach, even through the high school levels, in terms of access to this.
34:21 Chris Lehane
And then, from what I know, the public schools approach it one way, the private schools are it in an entirely different way in terms of actually teaching that literacy. Chuck, you used the word empower. I'm giving Kate a lot of really hard questions here, but what's the balance on the empowerment piece here? Because, like, if we get this right, ultimately, it is there to help kids.
34:46 Kate Ruane
Yeah, okay. Sure, I'll solve this.
34:48 Chris Lehane
Yeah. As I said, I'm giving you the hard questions here.
34:51 Kate Ruane
I mean, I certainly don't have the answer, but I would say that I've seen, you know, there's been programs in Europe where they have funded, you know, digital literacy education for kids from K all the way up through 12. And things like that are some of the things that I think about. I don't think that banning access to a particular technology, whether in school or out of school. And we've seen both proposals here in the United States is is helpful. And certainly outside of school, it's probably unconstitutional. So, you know, how do we how do we actually teach kids how to become experts while using this technology and become experts in the technology itself as as they age?
35:39 Kate Ruane
I think a school would be a good place to to begin to learn how to do that. And I also you know, I'm cognizant of the there was an op-ed by by a kid from Texas who where they have a policy that has banned him from him and other students from using this technology. And he is a computer science nerd and really wants to use it to build. And it's like the op-ed points out how how kind of silly this is from his perspective. So we have to. We have to be integrating support for kids education throughout, and all kids education throughout their throughout their life cycle.
36:18 Kate Ruane
And I don't think bans necessarily accomplished that. So so let's let's think about it a little bit differently and a little bit more supportive, in a more supportive way that that guides them along the process as we were hearing earlier from others on the panel.
36:36 Chris Lehane
Thank you.
36:37 Kate Ruane
Thank you.
36:38 Chris Lehane
Well, first of all, thank you everyone for being here. I think we could go on on this conversation for a while. It's obviously an incredibly important one. I would really have not asked for this permission from this august and distinguished group, but would really encourage folks here who are thinking about this and working on the policies. A lot of really smart people in the room. It reminds me of a, you know, a line when I worked way back when for for President Clinton, Chuck was in the Commerce Department and President Clinton used to come into rooms like this and he would say who's the smartest person in the room and there's always someone who raised their hand, by the way.
37:18 Chris Lehane
But the takeaway was that there's no one person. It's the collective brains in the room. And you know, on this panel, you have a lot of brainpower and I know in this room, a lot of folks who are giving an incredible depth of thought to how we think about this subject. So I would certainly encourage folks to reach out to the people on this panel and obviously, you know, we at OpenAI stand here, you know, really interested in being able to work with you all. Thank you for taking time from your busy day to spend with us and this is a conversation, but an ongoing conversation. So thank you very much.
+ Read More
Like
Comments (0)
Popular
Watch More
Event Replay: Sam Altman on Building the Future of AI
Posted Apr 06, 2026 | Views 10.9K
# OpenAI Leadership
# AI Governance
# AI Safety
# Economic Opportunity
Event Replay: Minus AI, Plus AI, Times AI — A Vision for an AI Pedagogy
Posted Nov 14, 2025 | Views 1.5K
# AI Education
# AI Pedagogy
# AI Adoption