Unpacking the Goals of Common Sense Media's Youth AI Safety Institute

Audio of this conversation is available via your favorite podcast service.

On May 5, Common Sense Media, the nonprofit known for its entertainment and technology recommendations for parents, launched its Youth AI Safety Institute, backed by a $20 million annual budget to “define what child-safe AI actually means” and to “rigorously test AI products” and assign them ratings.

The Institute is already working with policymakers in the US at the federal and state level, and is building initiatives abroad. On May 12, the Institute was introduced in Europe at the Copenhagen Summit, co-hosted with Save the Children Denmark and former European Commissioner for Competition Margrethe Vestager. Speakers at the event included European Commission president Ursula von der Leyen, Ofcom chief executive Melanie Dawes, and former US Secretary of State and presidential candidate Hillary Clinton.

The Youth Safety Institute’s funders include philanthropies and wealthy individuals, as well as companies such as Anthropic, the OpenAI Foundation and Pinterest. Common Sense Media's ratings have called ChatGPT-5 "high risk;” Grok, various AI toys and AI mental-health chatbots "unacceptable risk," Google Gemini’s K-12 product "high risk," and Claude "moderate risk."

The Youth Safety Institute will be led by Bruce Reed, who joined Common Sense Media as Head of Common Sense AI in March 2025 after serving as President Joe Biden's White House Deputy Chief of Staff, where Politico dubbed him the "AI Whisperer" for leading Biden’s AI Executive Order and securing voluntary commitments from frontier labs. Last year, Time named him one of the 100 most influential people in AI. Reed previously worked with Common Sense as a senior tech-policy adviser from 2015 to 2020, and was a lead negotiator on the 2018 California Consumer Privacy Act.

I caught up with Reed about how he views the current state of AI and child safety and his goals for the Institute.

What follows is a lightly edited transcript of the discussion.

Bruce Reed:

I'm Bruce Reed. I'm head of AI at Common Sense Media.

Justin Hendrix:

Bruce, I'm pleased to talk to you today and pleased to learn a little bit more about this extraordinary initiative that you are undertaking on behalf of Common Sense. I want to just step back a second and ask you a little bit about the world we're in right now. You have published research at Common Sense showing a large majority of children are already using AI regularly. More than half have interacted with AI companions, nearly a third say those conversations are as satisfying or more satisfying than talking to friends. How would you describe the current moment? These tools are widely distributed, lots of folks are engaging with them, already seeing harms pile up. What does the world look like at the moment that you're starting this new institute?

Bruce Reed:

Well, AI is a world of promise and peril. Kids have long been early adopters of new technologies and I think they've gotten way ahead of their parents and way ahead of politicians and policymakers and put themselves or allowed the companies to put them at considerable risk because one of the blessings and curses of AI is that we don't understand everything about it and that's allowing it to open new doors, but there are quite a few goblins behind some of those doors for these kids and teens. And part of the reason that we're so intent on making sure to do youth AI safety testing is that we think kids should have a good experience. They shouldn't be the crash dummies, the way frankly, they've been for social media. I think many in the industry look at what has happened to social media over the past decade and don't want to go down that road either.

But the other mistake we can't repeat with social media is on the part of the government and of watchdogs and we need to make sure that there are guardrails in place, but until that happens, we also need to make sure that parents and kids know what the risks are and can make a good judgment about what to do.

Justin Hendrix:

So I've heard that comparison to social media and the concern in particular that we're at great risk of making the same mistakes with AI that we made with social media. I think folks like former Surgeon General Vivek Murthy have said that. I know similar language from Common Sense's Jim Steyer. I guess one of the things I'm wondering about right now is whether we're still in this kind of preliminary period or whether we've already made the mistakes in a similar way that we made with social media. I mean, these tools now, as we're talking, hundreds of millions of people are using, have children are using across the world, in schools often being assigned and/or given the tools by school districts that in some cases are partnering up with the firms. Are we still in a sort of window where we're in a kind of preliminary place or are things already moving along?

Bruce Reed:

Well, that's a good point. I think we're in the process of making the same mistake. It's not too late and we should have learned our lesson by now, so I think one mistake that is definitely happening all over again is that Washington is largely absent from the scene. It remains an uphill battle, not impossible, but uphill battle in every state legislature to try to get some kind of guardrails enacted, even though protecting kids online and protecting kids from the harms of AI, is the most popular, most bipartisan issue in America at this time. It's a 90 to 10 deal everywhere, but there's a lot of pressure from the industry and sometimes from the administration to clear the road to move as fast as possible, and so we need to decide as a country and as a society that, "Yes, we're going to win the AI race. Yes, we can beat China. Yes, we can do amazing things with AI, but we don't need to harm, let alone sometimes kill our young people in the process."

Safe AI is going to make adoption easier, it's going to make progress easier. It's not going to slow down this technology. And I think the mistake the companies are making again is that most Americans look at AI through the lens of social media, and so that concerns them deeply. Most parents are terrified of what AI is going to do to their kids. At the same time, they're worried about what data centers are going to do to their communities, what AI is going to do to their job, down the road, what AI could do to their kids' education and to their kids' jobs. So it's not just important for the safety and sake of the kids, it's also going to be an essential prerequisite if we're really going to adopt AI as a society as opposed to reject it, set it aside or be miserable about it.

Justin Hendrix:

I want to ask you a little bit about the scope of the institute's work, what you plan to get up to. I know you're already making relationships both in state governments, federal government, you're working on policy agenda, but also abroad. I believe you've had some inroads in Europe and elsewhere. Can you just give us a sense of what it is you're building, what the organization looks like right now?

Bruce Reed:

We want to build the first Youth AI Safety Institute in the world. At the moment, there are a dozen national safety institutes that have been set up over the past three or four years that focus on national security risks, existential risks. They're doing very good work, but none of them focus on the immediate harms to kids. So there's a hunger for that in the United States and around the world. We just had a big summit in Copenhagen where the leader of the European Union came to endorse this idea and where there's a real hunger to protect kids. We think this is a global concern. We think that the Institute can set globally accepted standards and do testing that will reassure users and parents around the world and have America be setting an example for the rest of the world instead of the opposite.

Justin Hendrix:

One of the things that I wonder about is the extent to which you'll be able to apply some of the efforts that Common Sense has done in past, the kind of ratings and other approaches to research. It strikes me that this moment with AI, it's so complicated because AI products update so frequently, they behave differently in different contexts, sort of resist standardized testing in the way that perhaps other forms of media or other types of products or services might. I suppose that means you're going to need a pretty technical team. How many researchers will you hire? What will it look like? What do you imagine you'll have to do in order to set up a proper evaluation protocol?

Bruce Reed:

You're absolutely right. This is a big technical challenge. We feel like we're going to be able to attract a lot of good technical talent from the Bay Area where much of it is based and many young engineers are parents themselves and are quite concerned about this. I think what Common Sense Media brings to this is 20 years of experience, at first setting standards for what's age appropriate, what's safe for kids. And if you can get agreement on those standards, as Common Sense has done over the years, then you have a benchmark to test against and I think the technical challenge is not nearly as difficult as the standard setting challenge.

And one of the reasons that the industry can't do it for itself is that they're not trusted to set those standards of what's acceptable for kids and they're not good at it either. So you're right that AI is full of mysteries that in some cases the companies themselves don't understand as they're building it, but if we can set out tests that can be applied to the AI, then we'll be able to find the harms, we'll also provide a roadmap which the companies desperately need to try to prove that they're doing everything they can to make their products safe.

Justin Hendrix:

So one thing that obviously was kind of out of the gate of concern to some, is the fact that industry is involved in this, is providing funding to the effort. I suppose when you think about those other AI safety institutes, industry is often involved in those as well, but they are governmental or quasi-governmental organizations not typically quite so reliant on corporate funding at least in order to do the work they're doing. How do you think about that? How are you planning to ameliorate the appearance of conflict of interest or actual potential conflict of interest in the work that the Institute does?

Bruce Reed:

Well, Common Sense Media has been earning that independence for the last 20 years. We've done testing and ratings of social media products and other technical products and for that matter, TV and movies and so on, and we've been fiercely independent. We get yelled at all the time by the companies whose products we rate. That's not going to change. So our audience is the parents, they want a straight answer. We have a diverse mix of funding for the Institute from philanthropy as well as some companies. The companies have no say in the tests and the results, in what gets published, so it's going to be a fiercely independent institute and people will be able to judge it by its rigor. And it'd be great if government did this. It's a hard thing for government to do and current government's not in the mood to do it and tech money's influence in the political system is considerable.

So we think this is a great way to do this quickly and we've consistently held companies' feet to the fire on product. We've done risk assessments already that raise big safety concerns in some cases for Meta AI and safety concerns for ChatGPT and Claude and Gemini. So we call them as we see them and people can judge us by that.

Justin Hendrix:

I'm certain there'll be plenty of scrutiny to that and to any potential kind of divergence in your recommendations and where the companies sit. I know that's kind of already been an issue in some of the ballot fights and questions from the groups who may sit on the other side of some of those things. But let me ask you, when you do right now look out at the state picture in particular, there are dozens of states already thinking about legislation, thinking about ways to accomplish various AI safety goals when it comes to children, who do you think is doing it right? Are there any examples of legislation that you already think are in the right framework or ones that you think might be going in the wrong direction?

Bruce Reed:

Well, we've seen some real progress in the last few years. There's still a lot more to do. In California, were able to help pass a broader safety bill that was led by Scott Wiener, that it wasn't aimed at kids, but it's important of step forward on transparency for frontier AI models. We helped pass an age assurance bill that Buffy Wicks shepherded through the California legislature and I think we have a good chance this year to pass some strong chatbot guardrails in California. There are plenty of other states red and blue that are following similar playbooks and we've seen a lot of courage from governors, Republican and Democrat alike, to say they're not going to let Washington stop them. If the Trump administration doesn't care about youth AI safety, then they're going to prove that the states do.

So in an ideal world, we'd have national laws to do this kind of protection, but one thing we've learned over the years is that if we can pass a strong law in California where the companies are based, as we did in 2018 with the California Privacy Act, it can become a standard for the country and that certainly is far better than sitting on the sidelines and letting nothing happen.

Justin Hendrix:

So I sit in New York, so I have to ask you about what Governor Hochul's doing here. I know that you recently conducted an event with Governor Hochul. What are you up to in New York? What do you make of the governor's approach here? I know there's Safe for Kids Act, the AI Companion Law. Is New York's package potentially a template?

Bruce Reed:

Yes. She's been great on tech issues across the board. She was one of the first to go all in on getting cell phones out of the classroom, which is another important step forward for kids in tech. New York past a similar Frontier AI transparency and safety law last year that the governor signed. They've got good bill in the works on AI companions, so she's been a real leader and in the same way as it carries a lot of weight when California passes something, New York is the center of the media universe and so when something happens there, the world hears about it. So we have high hopes for New York as well.

Justin Hendrix:

Well, the idea of the ban that you brought up kind of prompts a question for me that I wonder about sometimes. I mean, you in your own ratings have called AI companions for kids chatbots for mental health support, you've classified them as an unacceptable risk for children and teens. If the evidence keeps accumulating in that direction when it comes to AI, I mean, I can imagine we'll see more longitudinal studies about impact on education outcomes, there are lots of worries about things like cognitive surrender and the idea that children will offload too much, or all of us, I suppose, adults as well, will offload too much to AI systems and that could have big implications for our own ability to acquire or process information or do various educational tasks. In that world, can you imagine the Institute arriving at a place where it is, for instance, potentially calling for outright bans in schools and things of that nature when it comes to this technology? Are those types of options on the table?

Bruce Reed:

Well, the Institute is not going to do advocacy. It's just going to play it straight and provide evidence and ratings of how this stuff works and the impact it has. So that's the Institute's mission. I think we will continue at Common Sense to push as hard as we can for strong guardrails on AI companions and other forms. I think it's very important to limit or prohibit a number of bad outcomes, whether there could be actual bans is more of a challenge with the courts, but I think one thing that the Safety Institute will help to do is to cast a spotlight on the harms that are happening and ideally the companies will fix those harms themselves, and if they don't, then it'll be up to parents and the legislative system to apply the guardrails.

Justin Hendrix:

Let me ask you, just cast your mind forward a little bit, because I always feel like to some extent the folks who run these types of initiatives, your imagination of the future matters to the shape of the initiative and where things might go, how the resources will be deployed. But I don't know, if you think forward five or 10 years, do you have a vision in your head for how children, teens can coexist alongside artificial intelligence in a safe way? What does that look like? Are there any tangibles in your mind? Are you exposed to conversations where people are trying to future or speculate or try to arrive at some sort of vision that's sort of different from the present?

Bruce Reed:

Well, I think that we certainly should be able to limit or eliminate the severe harms. The whole point of AI is to be as smart or smarter than humans and so we should be able to figure out, the company should be able to figure out a way to stop encouraging kids to commit suicide or do self-harm or any of the other harms that we're all worried about. So I think that, not that this is low hanging fruit by any means, it's challenging as the technology gets more powerful and that's true with plenty of other safety risks as well that aren't limited to kids.

But I think we can see a time when a parent could say that with appropriate parents' eyes over their shoulder, that a child could be able to use these products without fear of serious harm. I think a much more challenging question that we don't know the answer to and won't for a while, but we should worry about it and it will be one of the focuses of the Safety Institute over the long haul is, what does AI mean for education and what does it mean for learning and cognitive development or the opposite? And that's another thing that we got wrong with the first tech boom. We got laptops into the classroom and we got a lot of tech usage by kids, but it didn't lead to better outcomes, and there's definitely some evidence that cell phones and social media either contributed to or didn't stop the decline in reading over the last 10 years.

And I think the risks for AI are much greater than that. If we surveyed kids and parents about how they feel about AI in the schools and the parents were by a 20 point margin, concerned. They thought it was unethical to use AI in the schools. And the kids by a 20% margin thought it was a great idea and that seemed like a striking disagreement. But if you step back from it, the parent's concern was that AI was going to make the homework easier or solve the problems for the kids and the kids thought the same. It's just they were happy about it and the parents were alarmed. So I think we're going to pay special attention to the impact on education, which is not purely a safety issue, but in a way, if it's undermining kids' emotional and intellectual development, then that's a serious harm over the long haul.

Justin Hendrix:

Yeah. I think one of the interesting things to me is the extent that there's so much focus on how kids might use AI, a lot less focus on how teachers use AI or might use AI, and the ways that that might change the classroom and the educational experience as well, things like feedback and grading and assignments and all of the rest of that. I know I'm kind of witnessing in my own children points of evidence around those issues as well, which is really interesting to observe. But let me ask you just a couple of last questions. One on the level of engagement from the corporations you're working with now, who are you talking to inside those companies? I mean, at what level is the relationship held?

Bruce Reed:

We've been dealing directly with the top people at these companies and it's important that they be engaged from the outset, because the companies are going to have to do the lion share of the work of making this stuff safe. Our job is to show where they're getting it wrong and let parents know when the companies are getting it wrong and point to the risks and harms that the products have, and the executives are going to have a choice of whether they want to go rogue as some of the social media companies did and just to not care what the parents and the customers say, or whether they are going to do what most industries throughout our history have done, which is recognized that safety is a plus with consumers. We do safety testing for drugs, not just for kids, but for adults, to make sure they're safe.

We've for half a century done crash testing for automobiles and car makers learned after a while that actually it was a plus to have a safe car and we've seen a race to the top on safety with cars to the point that we now have the second-lowest traffic fatality rate in our history. So it's important to have pressure from independent organizations like ours and from parents and from other customers and ideally from government oversight to make sure that this industry does the right thing.

Justin Hendrix:

So you spent some decades in government, you had a very senior role, of course, in the Biden administration. You had a reputation of being a person who was essentially helping coordinate a response to AI as it came along. Has anything changed about your perspective on these issues or this technology since you left the White House? I know you're back and forth from Washington to Silicon Valley more often. Any kind of insight or anything that's changed your way of looking at things in the months since you've left?

Bruce Reed:

Well, I've been very encouraged by the bipartisan embrace of kids' AI safety at the state level and by how many governors from both parties have spoken out and how many attorneys general are holding the industry accountable. And so Washington has for a long time now been deeply partisan and struggled to get things done, but at the state level, it's still possible to make progress. But I learned something else while I was in government, which was that the same executives that the companies that we're talking to now recognized how fragile their enterprise is and what the broader risks are to society and world and national safety and the guardrails that we put in place on frontier models in the Biden administration were all guardrails that the companies voluntarily agreed to because they recognized that it made a lot of sense to test their products before they go on the market and it made a lot of sense to share those results.

And I think we've seen in recent weeks with Claude Mythos, that the administration has come to recognize the enormous national security, and for that matter domestic security challenges, that products as powerful as these frontier models can pose. And I think it's a good thing, not a bad thing if the national government is on the lookout for the national security and a kind of broader safety consequences of AI, because this is the most powerful technology that's ever been invented and we better have everybody on the lookout to make sure it's going to turn out okay.

Justin Hendrix:

We'll be paying close attention to the work of the Institute. I appreciate you joining me today, Bruce Reed. Wish you well.

Bruce Reed:

Thank you so much, Justin.