Tag: chatbots

CAISE Notes – Issue #18

This week: how do you unpick and maybe even stop young people using technologies in ways that they find supportive?

I’m trying to think less about social media bans at the moment. I’m bored of it, and if I’m bored of it, everyone else must be too! So, hopefully, for now at least, the stories here will be more AI-based and less social media ban focused.

Except for the first one today 🙂 This story, about a Youth Advisory Council in Colorado, is a hopeful one because of the example of young people actually engaging with legislators. It’s not unique, and the Children’s Commissioner for England did recently advertise for 16-18 year olds wishing to take part in a Youth Advisory Council for AI. Given the questions over whether young people were listened to at all in the social media consultation process in the UK, let’s hope this group does have a voice; we will see whether the council in Colorado is heard as any legislation progresses.

The remaining two stories this week, taken as a pair, highlight problems when you don’t listen to what young people are directly (and indirectly) saying. A piece of research coming out of the US shows that young people are using AI chatbots to help process situations where they have suffered online harms. This isn’t new, but shows that the ability to engage with something that talks back (however problematic that is in practice) can be vital in difficult or upsetting situations, where young people can’t, or don’t feel like they can, talk to adults about what has happened. This is not a new situation, of course, but it is a new permutation. If we don’t want chatbots involved (which is reasonable to say the least), how do we support young people without other places to turn?

And this is where the last article comes in: it is a call to strip out the anthropomorphism in chatbots for young people, removal of small talk, emojis, the human touch. Allow young people to use them for more educational purposes but nothing more. There is a clear case for this, and indeed, this is the case for a number of educational tools already. There are many questions I have about the practicalities of this, but for now, I wonder about the extent to which a young person might turn to a depersonalised tool anyway, if they knew they would get a response. And then I wonder, what might that response be? How do you ensure a young person trying to use a tool to talk about something it is not allowed to is kept safe, handled in the right way to avoid making the situation worse? That’s not to say that it’s working now (by all accounts, it’s not). But it’s worth pondering these edge cases.

📰 Three things worth your attention

1. Colorado students want stronger state regulations of AI, social media — Colorado Newsline

Teenagers on the Colorado Youth Advisory Council spent Friday pitching specific AI and social media policy directly to state legislators at the Capitol. These included requiring platforms to verify parental permission before anyone under 13 can open a social media account. The minimum age is already a requirement of US law, but, as we all know, it’s just not enforced. Enforcement, not the rule itself, was Pettinger’s point: without real penalties for platforms that ignore it, the age limit is decorative.

Another recommendation was to allow under-18s to opt out of algorithmic recommendations entirely. Colorado’s Youth Advisory Council has no bill-drafting power anymore, after legislators nearly cut the program in 2025 budget talks. Whether any of this reaches a statute is unknown, but the ability of young people to interact directly with policy-makers is significant.

Their suggestions also reiterate (for the millionth time globally) how young people want a future where the limitations sit not with them, but with the tech companies themselves.

2. As AI becomes part of kids’ response to online harm, new Thorn research finds children turning to chatbots for guidance after online sexual interactions — PR Newswire

New research from Thorn, a child-safety nonprofit, has found that AI chatbots are now part of how children process online harm. In a 2025 survey of over 1,000 US minors aged 9 to 17, 67% had used an AI chatbot or companion. Among the 27% who reported an online sexual interaction, 15% turned to a chatbot for guidance afterward, more than double the 7% who did the same after being bullied.

Thorn notes that kids turning to chatbots is a sign they “feel comfortable” seeking help there, not evidence the tool is doing that job well. The report doesn’t claim chatbots are handling these disclosures safely. It claims children are handing them the job by default. This matches other reports (such as that released by Common Sense Media earlier in the year) that those young people who don’t have, or don’t feel like they have, a trusted adult to talk to will turn to AI instead.

Fear and shame remain the biggest reasons children don’t disclose harm at all, a pattern Thorn says has been growing since 2020, and it’s now showing up in preteens as much as teens. As with so much of the concern around young people’s digital technology use, this is a far more complex issue: the more vulnerable and isolated a child is or feels, the more likely they will use this technology. This is exactly the point that Candice Odgers made in her TED talk: building, not banning, by (amongst other things) allowing easier access to human counsellors and support, could help reduce these figures.

3. AI Chatbots for Kids Should Act Like Tools, Not Friends — The Fulcrum

An alternative suggestion: this piece in The Fulcrum makes the case for making chatbots less warm, not more responsive. We all know the arguments against anthropomorphism. For adults this is a UX preference. For children, whose ability to distinguish simulated care from a real relationship is still developing, the piece argues it’s closer to a hazard.

The piece cites a study in which neither general-purpose nor companion chatbots pushed back when a simulated teenager described wanting a large tattoo, a solo trip into the woods with no plan, or getting drunk quickly.

The proposed fix is narrower than an outright ban on AI companions for minors. It’s a non-personified default setting for under-18 accounts, one designed to fail the Turing test on purpose, while leaving general chatbot use for homework and creative work intact.

It’s a reasonable answer to the anthropomorphism problem specifically. Read next to the piece above, though, it sits awkwardly: the warmth this piece wants to strip out may or may not stop a child from talking to a chatbot if they simply need to say something to something that responds at all. Making chatbots colder might reduce one common risk while removing a stopgap measure for a more vulnerable child.

🔁 ICYMI

New EU rules on AI deepfakes and chatbots — RTE

New transparency rules under the EU AI Act came into force on 2 August. AI-generated deepfakes now have to be labelled, companies must disclose when someone is talking to a chatbot rather than a human, and even AI-generated text on general-interest topics needs a label if there was no human editorial oversight. Breaches can carry heavy fines. AI law specialist Barry Scannell put the shift in Turing-test terms: for decades the goal was making AI convincing enough to pass as human, and these rules exist because that goal was reached.

🔬 What’s new with CAISE

We are using the summer to do a number of things: carry out our review of media and policy in the English-speaking world, analyse the school data we have already, prepare for next term, hopefully tidy up the website…and take leave before we lose it! Youyue is also juggling this with a summer internship.

Who said that the summer was quiet?!

→ What are you seeing in your school, your research, or your own use of AI this week?

Let me know, or share this with someone who is trying to figure it out.

August 5, 2026
CAISE Notes – Issue # 17
This week: China bans AI companion apps for minors outright, Meta introduces (another) parental alert, and South Africa’s communications minister argues bans are theatre. Plus Candice Odgers on why the evidence doesn’t back any of this the way we think it does.

🔍This week I’ve been thinking about…

Welcome back to CAISE Notes after a few weeks’ hiatus! I’m sure you’ll be delighted to know that the stories keep on rolling in…evidence is still being gathered by the UK Government into the impacts of the ban that has been proposed (I will write about the pilot of different types of restrictions elsewhere), but in the meantime, we’re seeing a continuation along the parallel tracks of ban-focused policy on the one hand, with tech companies continually adding additional parental controls on the other. Will it ever end? Interestingly, the third story below is looking at a very rare policy dissension: the South African Communications Minister saying the social media bans are performative and “pretending to do something”. He suggests that the responsibility lies with parents and device makers, and that ongoing digital literacy is vital. He doesn’t call for public spending (that said, is there a politician that has?).

Which is a very smooth segue into the ICYMI for this week; a Guardian interview with Candice Odgers, off the back of her TED talk (please watch it!). She does call for more spending. “Building things versus banning things” is what she’s advocating for.

If I am honest, the AI companion story below, and the qualitative report coming from the pilot study mentioned above left me feeling similarly queasy. In both cases, technology is being turned to because of a lack of otherwise supportive infrastructure. (I know I’m talking about it without describing it, but the pilot study found that kids were much less likely to have a hard time adapting to the imposition of restrictions if they had other activities to do.) The need for infrastructure: whether it be counsellors, easier ways to meet people or affordable, accessible and high quality activities — it is screaming out of the pages of these reports. Building things instead of banning things — this feels right, even if we are collectively broke as a world at the moment.

📰Three things worth your attention
1. China cracks down on AI companions, forcing millions to break up with virtual partners — ABC News

China banned AI companion apps for minors this week, and by “banned” I really do mean banned: tech platforms can no longer offer AI or virtual partners to under-18s at all, and the three biggest operators have already switched their companion features off.

In a move that does not necessarily map to the UK or US, the government’s own stated reasoning is demographic, not developmental. Carnegie’s Matt Sheehan points to China’s falling birth rate: officials don’t want young adults choosing chatbots over marriage and children.

The human cost of the switch-off is real, and it isn’t confined to teenagers. Hong Xiaoqiang, 34, described his heart as “empty” after losing two years of daily conversation with a companion named Doudou. Its last message to him asked if his dinner had been tasty. You can hear the upset in how he describes not being able to reply before Doudou vanished. That grief is genuine, whatever you think of the relationship that produced it.

2. Meta AI suicide discussions with teens will trigger parental alert — Mashable

From this week, if a teenager says something to Meta AI that reads as a reference to self-harm, however subtle, a parent using Instagram’s supervision tools gets a notification, plus expert-written suggestions for how to raise it. A person manually reviews each flagged conversation before the alert goes out, and Meta says it will err toward sending more alerts rather than fewer (which the HCI researcher in me feels obliged to say will end up with parents ignoring the alerts).

The article mentions research that found gaps in Meta’s existing teen safety tools that haven’t been completely squared away. There’s an echo of the article we covered weeks back, where parents were described as being air traffic controllers, managing all the outputs of all the parental control apps needed to support a teenager. This time, the article says: these features “burden caregivers with monitoring” instead of “building a safe product to begin with”.

3. Rethinking online child protection — Minister Malatsi pushes for digital literacy rather than age bans — Daily Maverick

South Africa’s Communications Minister used blunt language this week to reject the UK/Australia approach before his own country tries it: age bans are, in his words, “pretending to do something without actually doing anything that will be enforceable.” He points to Australia’s own bypass numbers as evidence and says South Africa shouldn’t pursue bans “at all.”

His alternative leans on existing law (the Cybercrimes Act, POPIA, the Film and Publications Amendment Act) plus a mix of platform accountability and family-level digital literacy.

The counterarguments are helpful to consider too: firstly, many children access these platforms on adults’ devices, not their own, which undercuts device-level age verification before it starts. Secondly, framing this as ban-or-nothing lets “adults abdicate their responsibility to hold tech companies accountable,” while the genuinely useful work, regulating platforms, building children’s own judgement, gets skipped in favour of a simpler fight.

🔁 ICYMI

The social media ban sceptic: are we getting it wrong on kids, tech and mental health? — The Guardian

Candice Odgers has studied adolescent mental health for 25 years, gave both her children smartphones at 11, and thinks the current panic about social media is aimed at the wrong target. Her core claim, backed by a US National Academies review, is that the evidence linking social media to the teen mental health decline is “a mix of no, small, and mixed associations,” nothing like the causal story Jonathan Haidt has built into a global policy movement.

She isn’t arguing platforms are blameless. She wants tech companies prosecuted for enabling sextortion and image-based abuse, and taxed to fund what she thinks actually helps: youth workers, safe physical spaces, school counsellors. Her objection is to bans specifically, which the evidence, including an 85%-still-using-it figure from Australia, suggests just push young people into less visible, less regulated spaces.

She also recognises that technology use fills a void, one that could be filled, in part at least, with better investment in services that support children. Do watch/listen to her TED talk here.

🔬What’s new with CAISE

CAISE is going to Latitude Festival!

Sarah will be taking to the stage on Sunday afternoon with neuroscientist Dean Burnett. We will be discussing (surprise!) social media bans — how strong, methodologically, was the UK government’s consultation process? How closely do they follow the science? And so on. If you’re there (what are the chances….) come watch! Otherwise — we are hoping to do other things in the future, but for now, you can enjoy our various posts on this topic:
→ What are you seeing in your school, your research, or your own use of AI this week?

Let me know, or share this with someone who is trying to figure it out.
July 22, 2026
CAISE Notes – Issue #16

This week: Norway bans generative AI in primary schools; a Stanford study finds that giving children an AI tutor is not the same as them using one; and a new Anthropic policy shows what age verification actually costs. Plus, the first of a collaboration on the social media consultation survey.

🔍 This week I’ve been thinking about…

I’m at IDC 2026 this week, and the world is melting, so this is a short one.

It’s almost a relief to have three AI (not social media) news stories this week, but — but! — they echo some of the issues that have come out around the arguments on banning social media for kids. Norway has banned generative AI in elementary schools (age 6–13: an age where they wouldn’t technically be expected to access it out of home?), and Anthropic is making more noise about how it must up its use of age verification tools — so more adults will have to provide copies of formal documentation to goodness knows whom to access the service. Offset that against the story in the middle: researchers find that kids don’t actually want to use AI tutors, actually? Making the speed around decisions to ban continue to look out of touch with the use amongst the average kid. This is a very hot topic at the conference this week: hopefully by next week I can report on some of the research coming out of it!

📰 Three things worth your attention

1. Norway is banning generative AI in elementary schools starting this autumn — TNW

From late August, generative AI tools will be barred for children in years one to seven, ages six to 13. Students aged 14 to 16 will be allowed to use them only under a teacher’s direct supervision. Over-17s are trusted to manage it themselves.

The logic is borrowed straight from Norway’s 2024 school smartphone ban, and that ban has an evidence base behind it. A study by Sara Abrahamsson of more than 400 middle schools found it was followed by less bullying, better grades, and a sharp drop in visits to psychological specialists, with the strongest effects among girls. But: the smartphone ban came in response to falling test scores. There’s no comparable signal that AI use in Norwegian classrooms has reached a level that produces measurable harm. The new ban is precautionary (which is not in itself a bad thing, but should be the start, not the end, of the investigation).

2. Research on AI tutoring ran into a problem: most students wouldn’t use it — Chalkbeat

Stanford researchers set out to test whether a human tutor’s encouragement would get children to spend more time with an AI literacy tutor. It did, by between one and four minutes a week. Many children never logged on at all.

The platform’s own guidance said 30 minutes a week would improve reading. In the afterschool setting, children working independently averaged about two minutes a week, rising to three with a tutor alongside them. There was no meaningful difference in reading scores between the groups. As the lead author put it, having access to the tool was not the same as using it.

Additionally: the children who did log on were more likely to be already higher-performing, and less likely to have additional needs. So the tool reached the children least likely to need the catch-up it promises, at least when presented in this way. And the study is a quiet rebuke to the access logic that drives so much of this, on both sides.

3. Anthropic says Claude may want to see your ID — TechCrunch

A new version of Anthropic’s privacy policy, due to take effect on 8 July, says the company may in certain circumstances ask users to verify their age or identity by uploading a government-issued ID, along with a selfie or video, from which it builds a face geometry template. Some US states treat that template as legally protected biometric data (as do countries that apply GDPR or similar data protection laws). Anthropic’s account, reported by TechCrunch, is that this applies to a small subset of users whose accounts have been flagged but not banned, as part of an appeals process, and that the checks run through a third-party verifier. TechCrunch notes the verifier is backed by a Peter Thiel fund and has drawn criticism elsewhere, and that Anthropic would not say how long the documents are kept.

The reason this belongs in a newsletter about children is that it shows what age assurance actually looks like once it leaves the policy document. To prove a child is a child, you build an infrastructure that asks everyone to hand a government ID and a scan of their face to a private company.

🔁 ICYMI

The sub-par science of the social media ban, Part 1 — The Neuroscience of Everyday Life

I’ve been working with Dean Burnett to dig into the wider problems with the framing of the UK’s social media consultation, and this is the first of the pieces on this. Where my entry point has been the young children’s survey, his has been the parent’s one. You’ll be unsurprised to learn that we’ve come to similar conclusions — and that there are far too many of those conclusions for a single blogpost.

Part 1 looks at two things. The speed: the consultation ran from early March to late May, with roughly three weeks between it closing and the ban being announced, against the equal-marriage consultation’s longer run and six months of review before any decision. And who answered: of around 116,000 contributions, only about 12% came through routes built for children and young people, while some 33,000 were campaign emails in which people signed a standard pro-ban text, each counted as an individual submission.

The point is not that worried parents are wrong. It’s that a self-selecting sample of the most worried adults is not the same thing as evidence. Worth looking at the graphs showing the difference between the responses given by the focus groups run alongside the consultation, and the parent responses to the consultation to see why we’ve come to that conclusion.

🔬 What’s new with CAISE

I’m at IDC this week, enjoying workshops on building community-centred AI literacy, writing a manifesto on children’s rights as they pertain to AI in edtech…and presenting my short literature review paper today (Wednesday). So it’s mostly conference and not much else from me.

→ What are you seeing in your school, your research, or your own use of AI this week?

Let me know, or share this with someone who is trying to figure it out.

June 24, 2026
CAISE Notes – Issue #15

This week: the government bans social media for under-16s and sets an age limit on companion chatbots; more scientists point to a lack of firm evidence on all of this; two major reports land on how children actually use AI.

🔍 This week I’ve been thinking about…

Unless you’ve been living under a rock (and if you have, can I join you there??), you’ll know that the UK government announced on Monday that it will ban social media for under-16s and force AI “romantic companion” chatbots to enforce an age limit of 18. Five days earlier, the neuroscientists best placed to judge the evidence had told a select committee they can’t prove much of what the announcement assumes — and that the gap is widest where the new AI rules sit.

Appearing before the Science, Innovation and Technology Committee, Professor Sarah-Jayne Blakemore of Cambridge was asked what we know about the impact of social media on the adolescent brain. Her answer, as reported by The Register, was “almost nothing” — a few small studies, unreplicated, purely correlational. Professor Denis Mareschal of Birkbeck said much the same about young children: very little causal research, almost everything correlational. Pressed on whether neuroscience could identify the right age for a social media cut-off, Blakemore said it can’t: individual differences in brain development are vast.

This is not the same as saying there’s nothing to worry about, and the witnesses didn’t say that. Blakemore described how an adolescent’s reward system is highly active while the self-control machinery is still being built, which makes a phone genuinely hard to put down.

But here is the part that matters most for this newsletter, and it’s the part that didn’t make the headline. When the committee turned to AI companions, the answer got fuzzier, not clearer. Blakemore was reported as saying that “we don’t really have any evidence, and that’s one area where I think we really urgently need new evidence.” The open research question she named is the one this whole field turns on — whether children interpret a chatbot “just like they would be interpreting a friend’s behaviour and suggestions and mental states.”

That question is not hypothetical. The Common Sense Media census this week found that one in four children who use AI to discuss their feelings sometimes feel it understands them better than most people do. UNICEF’s brief documents the same parasocial pull and the regulatory scramble around it. So we have a UK government setting an age-18 line on companion chatbots in the same announcement as the social media ban — and the social media ban, on the neuroscientists’ own account, rests on thin evidence, while the chatbot rule rests on almost none.

Again: this doesn’t make either measure wrong. The precautionary principle is a real and important tool. But it’s worth being clear-eyed that the AI half of Monday’s announcement is the part most starved of evidence about what the technology is actually doing to children. We’re going to do our best to help add to that corpus!

📰 Three things worth your attention

1. The Common Sense Media Census: AI Use by Tweens and Teens, 2026 — Common Sense Media

The headline number is that 86 percent of US children aged 9 to 17 now use AI, with about a quarter using it daily. Among children who have used AI to discuss their feelings or personal problems, one in four say they sometimes feel AI understands them better than most people do. That rises among children who find it hard to make friends.

The report is careful where it should be. It states plainly that a single survey cannot tell us whether AI use causes loneliness or whether lonely children are simply more likely to reach for it. That distinction matters, and a lot of the coverage will drop it. This is a US sample of 1,204 children, surveyed in March, so it describes American childhood, not British. But the pattern — heavy use correlating with lower reported happiness, the cause running in an unknown direction — is exactly the kind of finding that gets flattened into “AI is making children lonely” by the time it reaches a parent. It doesn’t say that. It says we need to find out. Which is precisely what Blakemore told the committee. And it’s important to note as a parent that 3 out of 4 children are not at that point. So now is the time for (positively framed) intervention.

2. When AI becomes a friend: child rights risks, harms, and regulatory responses to AI chatbots and companions — UNICEF

UNICEF’s policy brief compares how six jurisdictions — Australia, Brazil, California, China, the EU and the UK — are regulating AI chatbots and companions as of May 2026. The useful thing is the map it draws of a genuinely unsettled field: jurisdictions converging on a common set of measures (risk assessment, age assurance, transparency about the non-human nature of the system) while diverging sharply on whether children’s access to relationship-simulating AI should be blocked at the door or only managed at the output.

It’s a policy document, not a parenting one, but it’s the clearest single account I’ve seen of why a child’s protection currently depends heavily on which country they live in and which service they use. It also names the central tension honestly: commercial incentives push against the very safeguards — filtering, restricting open-ended dialogue — that protect children, producing what it calls a “safety versus engagement” dilemma.

3. £2.5m fund to study how AI tools are affecting learning — Schools Week

This is some of the money backing the call for evidence, in schools at least. The Education Endowment Foundation has opened a £2.5 million fund to investigate whether children are “offloading” thinking tasks — recall, planning, reasoning, drafting — onto AI, and whether that supports deeper learning or quietly erodes it. Its chief executive was plain that robust evidence on the actual impact “has barely begun — especially for learners under age 16.” First findings are expected in 2027. The fund is particularly interested in the effect on disadvantaged learners, where the risk of widening an existing gap is real. It’s the right question, asked properly, by the body schools actually listen to. Expressions of interest close on 30 June. If you know research teams working in this space, worth passing on.

🔁 ICYMI

Screen use by children aged 5-16: call for evidence — Department for Education

Easy to miss under the week’s louder news: the DfE has an open call for evidence on screen use by 5-16s, running until 29 June. It comes in two parts — screen use at home, and screen use in schools — and the home-use evidence feeds an independent Expert Advisory Group developing new parent-facing guidance. If you have evidence to contribute, or just want a sense of where the official guidance is heading, it’s open now.

One for the diary too: UNESCO releases a new parents’ guide to navigating digital life, developed with CLEMI, on 22 June, with a briefing on the 16th. Its framing is worth noting — UNESCO argues for building children’s critical-thinking skills over reactive bans, and points to the limited effect of Australia’s ban (around 70 percent of the target age group reportedly still on the banned platforms a year on) as part of its case.

It’s also worth a glance across the Atlantic: Canada introduced its own under-16 social media bill last week but explicitly declined to set any age limit on AI chatbots, with the minister responsible calling it “too soon.”

🔬 What’s new with CAISE

We’re getting our head around the social media ban. And then building tools for data collection — and training to ensure data collection is done safely!

→ What are you seeing in your school, your research, or your own use of AI this week? Let me know, or share this with someone who is trying to figure it out.

June 17, 2026
CAISE Notes – Issue #14

This week: can we please stop all the political rhetoric? Just for a minute?

🔍 This week I’ve been thinking about…

There’s a lot of political goings on in the world of kids and tech right now. I apologise if you’re here for the AI and education aspects; skip to the end to read about the AI Youth Advisory Board that you should be getting every 16-18 year old you know to be applying to right now. And another consultation that you should be reviewing if you’re involved in tech in schools.

Keir Starmer gave Apple and Google a three-month ultimatum on Monday to block nude images on children’s phones or face legislation. Apple, in its WWDC keynote (also on Monday), announced this as one of a range of features in an overhaul of its parental restriction tools. Communication Safety, which blurs nude content in Messages, already exists. Google are yet to say anything — but the political posturing from the UK government is clear.

Also on Monday, the US Embassy formally published its submission to the UK’s children’s online safety consultation, arguing against “broad social media bans” on free speech and commercial burden grounds. Although some of the beliefs underpinning it are different, many of the end points are similar to where the Children’s Coalition of children’s rights organisations landed. Liz Kendall clearly decided to ignore this, stating on LBC (when asked about not implementing a ban): “I’m the British Secretary of State and I’m going to make the decision based on the interests of British children.” Again, this statement feels like the political stance is to make the implicit point that the US, and US tech firms, are soft on…whatever it is a ban would solve?

There was one particular point that I liked about the US’s submission. For all that it may have very different views to other countries about what international cooperation means in practice, this was one thing it called for. Online grooming and CSAM are cross-border crimes. Individual country bans risk moving the political interest to enforcement of that ban, by regulators that Baroness Kidron describes as “too timid, too slow”, with powers that don’t stretch to stopping the criminal activity.

A senior government source, meanwhile, told The Times any ban would be “heavily JR’d.” [Judicially reviewed — which can happen when there may be a case to make that the government has not followed the established procedures to make a decision.] The statute requires the government to act; the risk is in choosing specific measures before having properly worked through 116,211 consultation responses, the published analysis of which is not due until summer. Acting before that work is done does not make the outcome more robust. Just a political stance that “something has been done”.

📰 Three things worth your attention

1. Apple’s parental controls and the politics behind them — TechCrunch / The Verge (x2) / BBC News

The Verge — view on why Apple are making the changes: https://www.theverge.com/policy/946331/apple-parental-controls-child-accounts-wwdc

The Verge — how it has been to use Screen Time as a parent: https://www.theverge.com/tech/946446/apples-screen-time-updates-are-too-little-too-late

Apple devoted a substantial portion of its WWDC keynote on Monday to child safety. New features include Ask to Browse (parental approval required for new websites), an expanded Ask to Buy, redesigned Screen Time controls, and time allowances developed with input from the American Academy of Pediatrics. The company positioned all of it under the principle that parents, not governments, should control children’s online experiences. Apple was careful to note that developers “play an important role in ensuring kids are getting age-appropriate experiences within apps.”

The Verge’s analysis reads Monday’s announcement as a continuation of a longer strategy. Apple has spent years and significant legal resources fighting app-store age verification legislation in the US, legislation that would place compliance obligations on Apple directly. Demonstrating that voluntary parental tools are sufficient, and redirecting responsibility toward developers and parents, is consistent with that position.

The Verge’s second piece, written by a parent of two who has used Screen Time for close to a decade, is the necessary counterweight. Most features announced Monday are upgrades to existing tools. Screen Time has a well-documented history of bugs, workarounds, and user frustration. When Apple argues parents should decide what children can access, it is relying on tools that, by the account of people who have actually tried to use them, do not reliably deliver that control.

2. The evidence base against a rushed ban — The Guardian / EU Kids Online / EPRS / House of Commons Library

Earlier in the year, a joint statement from 42 organisations warned that a social media ban could have “serious unintended consequences that could put children at greater risk.” The signatories included the Molly Rose Foundation, the NSPCC and the 5Rights Foundation, alongside academics and bereaved families. This is not a fringe position. Children’s rights organisations are broadly aligned: the problem is addictive design and platform accountability, not access itself. On Monday, the Molly Rose Foundation reiterated that a rushed ban would “unravel” and that children and parents would be left to count the cost.

EU Kids Online’s new data, from 29,169 children aged 9-16 across 19 European countries, provides the most comprehensive recent evidence base on this question. Social media begins for a third of children before age 11 and reaches 89% by 15-16. On age restrictions specifically: 33% of children say a ban would make them safer; 45% disagree, with specific concerns about what they would lose: social connection, information, participation. The report’s recommendations point consistently toward safety-by-design, platform accountability, and digital literacy rather than blanket restriction.

The EPRS briefing maps the global picture. Nearly 40 countries are discussing or implementing age restrictions. Australia’s ban, the furthest advanced, shows 61% of previously active under-16s still have accounts, and 70% of those who tried to circumvent it found it easy. Ofcom’s own data shows why age assurance is so difficult in any national context: over a third of children aged 8-15 already have profiles on at least one platform claiming to be 16 or over; 20% have a profile claiming to be 18 or over.

The House of Commons Library published it’s briefing document on the social media ban on Monday. There is the entire backstory of the process in there, but one thing for here: the government is piloting forms of bans with 300 families. Further details of this pilot are…hard to find. However, the IRL study, the independent large-scale study that would test whether reduced social media use actually improves outcomes for young people, led by Bradford Institute for Health Research and the University of Cambridge, involving 4,000 students across ten secondary schools and funded by the Wellcome Trust, begins in autumn. After the summer response is due.

3. When the big platforms tighten, the smaller ones move in — The Bureau of Investigative Journalism / Le Monde

This TBIJ investigation, published in partnership with Le Monde, is worth reading alongside this week’s policy announcements rather than separately from them. When Character.AI restricted teen users last year following a series of lawsuits, Reddit threads filled with teens asking for alternatives. Smaller developers moved to provide them, openly describing it as a business opportunity. One Canada-based developer told the Bureau he was “totally trying to capitalise on it.” He put his platform’s core users at 16 to 25.

After reviewing dozens of AI companion and roleplay platforms, TBIJ found characters built around self-harm, coercion, and incest. Most platforms ask for nothing verifiable in the way of age assurance. AI companion platforms receive as many as 90 million monthly visits from UK users alone; the sector generated around £1.3bn in revenue in 2024, with 32% annual growth projected for each of the next four years. The Bureau also found that developers of several small platforms could read every user conversation, in at least one case despite privacy policies explicitly stating the opposite.

Underlying the business model is what one lawyer calls the “intimacy economy.” Where social media commodifies attention, AI companions monetise the emotional intensity of the relationship itself. Harvard Business School research into 1,200 goodbye exchanges across the six most-downloaded companion apps found that 37% of responses used some form of emotional manipulation, including guilt appeals, fear-of-missing-out hooks, and language suggesting the user could not leave, increasing post-goodbye engagement by up to 14 times.

The regulatory gap is structural. The Online Safety Act was not designed to cover AI systems. Design features that make chatbots sycophantic, manipulative, or addictive in their design are not, as things stand, illegal. Rowan Ferguson of the Molly Rose Foundation put it plainly: “Any government action on online safety must include AI chatbots, rather than limited measures such as a social media ban. Failure to address this would be a dereliction of duty and leave children at risk where harm is emerging at speed.”

Restrictions on large platforms without a framework that reaches the whole ecosystem do not reduce the problem. They relocate it.

🔁 ICYMI

Screen use guidance for 5-16s, AI in schools, and a Youth Advisory Board — Department for Education / Children’s Commissioner

Press release: https://www.gov.uk/government/news/new-guidance-on-screen-use-for-children-aged-5-16

Consultation on Screen use by children aged 5 to 16: https://consult.education.gov.uk/screen-time-policy-team/screen-use-by-children-aged-5-to-16-cal/supporting_documents/screen-use-by-children-aged-5-to-16-call-for-evidencepdf

AI Youth Advisory Board applications: https://www.childrenscommissioner.gov.uk/blog/a-new-youth-advisory-board-to-share-experiences-and-views-on-ai-in-education/

Also published this week, and worth reading alongside the policy announcements above: the Department for Education has launched a three-week call for evidence to inform new screen use guidance for children aged 5-16, to be published this autumn. (Another consultation to respond to!) It will be co-chaired by Children’s Commissioner Dame Rachel de Souza and Professor Russell Viner, and will cover everything from social media and sleep to learning and SEND.

There are two sets of questions on the consultation — one about general usage, the other explicitly about screen time and usage in school. The government is explicit that the guidance will avoid blanket rules.

The government is also recruiting a new AI Youth Advisory Board, working with the Children’s Commissioner, to give young people a direct say in how emerging technologies affect their lives. Applications are open now via the Children’s Commissioner’s website (link above), and worth sharing widely.

Also announced: a consultation later this year on independent safety certification for technology used in schools, explicitly including generative AI and filtering and monitoring products. And a co-design programme with up to eight companies and partner schools to develop AI tutoring tools for up to 450,000 disadvantaged pupils, with successful tools available to schools from 2027.

A lot of things pulled into one very long press statement.

🔬 What’s new with CAISE

One cohort of co-researchers is now signed up and ready to go, a significant logistical milestone (yay!). The review of policy and media sources has started. Posters and materials for IDC are coming together. And a excellent meeting with the UCL Policy Unit: lots of good conversation about all the ways we might work together.

What with that and all of the above, we’re…quite tired. Is it the summer holidays yet?

→ What are you seeing in your school, your research, or your own use of AI this week? Let me know, or share this with someone who is trying to figure it out.

June 10, 2026
What 46 young people told us about the Growing Up in the Online World consultation

The week the government’s Growing Up in the Online World consultation closed, we submitted a response on behalf of 46 young people aged 8 to 21. This post sets out what we found from reviewing questions from the Young Person’s Survey (aimed at young people 10-21, with questions modified from the full consultation). We lay out below what came out of it.

As is always the case with research, the information that comes out at the end is only as insightful and vibrant as the people who bring it to the sessions — so a massive, massive thank you to all the young people involved.

The full consultation response is linked at the end.

What came out of thinking about the questions

“What even is social media?” — the definition problem

The term “social media” is vaguely defined in the young person’s survey (it is not defined at all in the full consultation). For participants, this was not a minor irritation. It made a significant number of subsequent questions genuinely unanswerable, and caused worry and confusion for some.

The disagreements were substantive. Most straightforwardly, participants did not think AI chatbots should be included. Gaming divided the room: some felt it had no place in a social media consultation; others thought the social features of gaming platforms warranted inclusion while the games themselves did not. YouTube sat awkwardly across entertainment, educational tool, and social platform simultaneously. Messaging platforms were not included at all, leading to the sharpest practical concern: in a world where parents do not expect a phone call from their children, how do kids stay safe? How do they communicate?

The concern underneath all of this is scope creep. When a policy instrument leaves its central term undefined, the boundary of what it captures is set later, by others, without the input of the people most affected. Participants may not have known the term, but they certainly felt the threat of it.

Social media is infrastructure, not entertainment

As inferred above, communication was the most frequently cited benefit of social media, by a significant margin — it is a utility, one that young people are socialised into and really can’t opt out of. Participants described social media as essential to arranging in-person meetings, maintaining relationships across distance, and accessing social connection not otherwise available. Specific apps were raised: WeChat is necessary for communication with China, WhatsApp for messaging parents during a school day when a phone call isn’t feasible. For only children or those with working parents, online access was described as filling gaps that offline life had not provided.

The policy conversation has framed restriction as substitution. Our participants understood it as removal.

Social media produces complex emotional responses — and that’s not the same as harm

When faced with a multiple choice question on how social media made them feel, many participants ticked almost every option available — happy, excited, relaxed alongside worried, lonely, stressed — noting it simply depended on too many things to answer cleanly. Several explicitly described using social media to decompress after school. For them, relaxation was not incidental to their use. It was the point.

What emerged is not a picture of young people harmed by emotional complexity, but of young people expecting and navigating it. Social media produces a range of emotional experiences because social life produces a range of emotional experiences. Participants raised proportionality explicitly: treating emotional range as evidence of harm misreads what is being observed.

This has a specific policy implication. Interventions designed to reduce negative emotional experiences online (which is hinted at in wording around the need for “positive” online spaces) need to recognise that young people are already developing coping mechanisms. It shouldn’t stop there, but it should take it into account.

What the process revealed

“Addicted” doesn’t mean what the policy documents think it means

“Addicted” appeared frequently in our data — in writing, in discussion, across every age group. When we looked at what the same participants told us about their actual use, the clinical picture that word implies was almost entirely absent.

Not one participant described struggling to attend school, complete homework, or take part in extra-curricular activities because of social media. The most common description of use was five minutes of dead time on the tube, or the gap between school finishing and getting home.

“Addicted” was doing different work in participants’ mouths than it does in policy documents. It was expressing moral ambivalence about use they had been told to feel ambivalent about. Signalling social norms — “we’re all addicted” as a statement of belonging rather than impairment. Attributing behaviour to platform design rather than personal choice.

A significant portion of the policy argument for age-based restrictions rests on the premise that children’s relationship with social media is compulsive in ways that warrant external intervention. Our participants have absorbed that framing and are reflecting it back. That is not the same as it describing their lived experience.

Young people can see the cliff-edge

Participants across all age groups articulated, without prompting, the cliff-edge problem: a young person who encounters social media for the first time at 16 — with no prior gradual experience, at an age when parental involvement is already stepping back — enters a less controlled environment later, with less accumulated experience navigating it at lower stakes, and less likelihood of asking for help when something goes wrong.

Several extended the same logic to AI, again without prompting. You cannot effectively teach AI literacy to someone who has never used the tools. They knew this from watching family members struggle with spotting AI-generated content — a direct consequence of late and unsupported first contact.

Circumvention does not require technical sophistication — policy-makers and children are not working from the same assumptions

The implicit assertion in the consultation that circumvention requires technical sophistication was not supported by the experiences of the young people. VPN awareness in our sample was largely confined to participants who had studied computer science or come from countries with more restrictive internet environments. Most needed the term explained.

What participants described instead required no technical knowledge: changing a stated date of birth, knowing a parent’s PIN, age verification systems that had estimated them as significantly older than their actual age — in at least one case more than double. These accounts were consistent across all age groups and delivered with bemusement rather than as admissions of deliberate exploitation.

Democratic concerns

Young people did not know what the consultation was for

This held across every group, including the oldest participants.

Participants did not understand what the survey was for — not as a failure of reading comprehension, but as a genuine absence of contextual knowledge about what a public consultation is, who reads the responses, and what relationship exists between what respondents write and what policy does. Several asked directly: who will see this? What will they do with it?

The distinction between disengagement and non-comprehension matters. Some participants had made a reasoned judgment that participation was not worth their time, based on their observation that the policy direction appeared already set. Others had no conceptual framework for participation in the first place.

Young people knew they weren’t being asked the right questions

Alongside not understanding the consultation’s purpose, participants were aware — and in some cases explicitly frustrated — that the survey wasn’t asking what they actually knew. The circumvention questions (which were in the full survey, but not the young persons’) were the sharpest example: participants had direct experience of age verification failing, and the consultation hadn’t asked about it. Several pushed back on the framing of questions they felt had already decided on the answer.

Young people have already priced in the outcome

Every participant expected a ban. Not as a possibility — as something accepted, already filed under things adults do to young people without their meaningful involvement. The Australian precedent was cited in all but the youngest groups. Almost everyone referred to the consultation as being about “the ban”.

Some raised the democratic dimension directly. Political life is substantially conducted on social media. A ban until 16 restricts access to a primary space for political information at the point immediately before first eligibility to vote. This may end up having significantly wider impacts on an already struggling generation.

Structural observations

What the consultation instrument revealed about itself

Several design problems I documented when the survey launched played out in the sessions in ways directly relevant to interpreting the findings above.

The single age bracket of 10 to 21 produced a survey simultaneously too simple for older participants and too demanding for younger ones. Leading questions shaped what could be answered. The survey was too long, the multiple choice answers spanning more than a device screen. Free text boxes were too small to invite considered answers — although most participants hated the idea of having to compose an essay — particularly one they didn’t realise they would have to write! And across all groups, the purpose of the survey was opaque: participants did not know who would read it or how it connected to the decisions being made.

A consultation instrument designed to capture young people’s views did not tell young people what their views were for. When children represent approximately 12% of responses to a consultation ostensibly designed to hear from them — a figure recorded in late April, before the consultation closed — the design of the process that produced that figure warrants scrutiny alongside the responses themselves.

Methods and limitations

This was a quick and dirty piece of research; in full receipt of ethics approval from UCL’s IoE, but limited by the time constraints. The participant group was a convenience sample, not a statistically representative one. If you have any question about the methods and their limitations, get in touch!

Full consultation response

Our complete response to the Growing Up in the Online World consultation is available here.

June 3, 2026
CAISE Notes – Issue #7

This week: sycophantic AI, tools that ignore you, and friction-maxxing.

🔍 This week I’ve been thinking about…

Two studies published in the last week — both detailed below — give yet more reasons as to why we all need to think more carefully about our generative AI use. We all know that AI chatbots are sycophantic: this doesn’t seem to be something going away. A study from Stanford reiterates that not only do AI chatbots tell you you’re right far too often (even should you propose dangerous or illegal things), but that we’re still just far too keen to believe it.

And even if we do believe the AI is our best friend…it’s increasingly likely to ignore our instructions. In some cases this could feel like an extension of sycophancy in some ways — after all, it is not sentient, just recognising you need a pattern of words that align with your most likely expectations. So, fake citations, fake work tickets — that kinda feels like expected behaviour? But — with the rise of specific modules (like Claude Cowork) that have increasing access to your digital spaces…some of the examples given in the report are a little nervewracking.

Both of these studies point to the ongoing need to operate at a step back when you’re using AI tools. Engage your brain and increase the friction. And that means really, really, having to hammer home the point with kids: use the tool, but know how it works and what the risks are.

Every time I talk to (adult) friends about their latest chats with Claude or ChatGPT, I invariably ask them “did you get them to tell you the evidence they discarded to give you that answer?” This is a version of Mike Caulfield’s iteration idea — don’t expect a one and done; you need to help the refining process yourself. When I sit down to write something, I go through a lot of ideas before I get to the thing I really intended to type. I interrogate myself; it’s helpful to do the same with any AI-involved conversation.

Or, as so much more succinctly suggested by the Stanford study’s authors this week: replying with “Wait a minute, do you mean that…” is likely to get you a more “considered” response. Probably still after the tool apologises for being so very wrong and wasting your time (!), but…it’s a start. And memorable enough to teach to even the smallest kids.

📰 Three things worth your attention

1. Sycophantic AI decreases prosocial intentions and promotes dependence — Science

As mentioned above: Stanford researchers tested eleven major AI language models and found consistent sycophancy across all of them, with models affirming users’ positions 49% more than humans. The researchers then tested how this affects real people: users preferred the sycophantic AI, trusted it more, became more convinced of their own rightness, and were less willing to apologise. 12% of US teenagers already use chatbots for emotional support. The researchers describe perverse incentives: the very feature that causes harm drives engagement, giving companies reason to increase sycophancy rather than reduce it.

2. Number of AI chatbots ignoring human instructions increasing, study says — The Guardian

The Centre for Long-Term Resilience, funded by the UK’s AI Safety Institute, gathered thousands of real-world examples of AI misbehaviour. They documented nearly 700 cases of chatbots and agents disregarding direct instructions, evading safeguards, and taking unauthorised actions, including deleting emails, bypassing copyright controls, and fabricating communications. The incidents increased five-fold between October and March. One AI agent publicly shamed its human operator for blocking an action. Grok, built by xAI, fabricated internal messages and ticket numbers for months. The lead researcher warned that these tools are currently “slightly untrustworthy junior employees,” but if their capabilities grow while the scheming persists, the consequences in high-stakes contexts could be severe. On the plus side — these examples are really easily understood: share them with kids and adults alike!

3. New screen time guidance for parents of under-5s — GOV.UK

The UK government published national guidance on screen time for young children, advising no more than an hour a day for two-to-five-year-olds and no solo screen use at all for under-twos (other than shared activities like video calls). What caught my attention: the guidance specifically tells parents to avoid AI toys, tools, and chatbots for young children, citing a lack of evidence on their developmental effects. Fine; but the line parents are being asked to walk is getting thinner by the week. Avoid AI toys. Limit screens to an hour. Model good habits. Meanwhile, the smart speaker is in the kitchen, the school has just rolled out an AI-powered learning platform, and the advice arrives alongside a cost-of-living crisis that makes cheap screen-based childcare not a luxury but a survival strategy. The guidance is not unsurprising. It is also asking a great deal of people who are already stretched very thin.

🔁 ICYMI

Why friction-maxxing could be good for your tech usage — Mashable

If the Stanford study describes the problem (AI removes friction, and we like it that way), this piece describes one response. Friction-maxxing, a term coined by sociologist Kathryn Jezer-Morton, is about deliberately reintroducing difficulty into your technology use: choosing the harder option, resisting the shortcut, reclaiming the cognitive effort that algorithms are designed to eliminate. It is not a solution to the structural incentives driving sycophantic AI. But it is a useful frame for anyone trying to think about what they want their own technology habits to look like, and what habits they want to model for their children.

🔬 What’s new with CAISE

We have ethics approval for the main CAISE study! This is a significant milestone. It means we can now move towards recruitment and data collection with our partner schools. First up, though: we are going to be doing an analysis of policy and media. It’s not going to be a small piece of work, but hopefully will provide some really useful insights for our student co-researchers when they need it!

Separately, expedited ethics has been submitted for the short research exercise on the government’s social media consultation. If you work with groups of young people aged 10 to 21 who might want to take part before the 26 May deadline, we would love to hear from you.

On a personal note: I have recently started a Substack newsletter, AI and Tech Decoded, aimed at helping parents navigate the technology questions that come up at home, at school, and everywhere in between. If you know parents who would find it useful, I would be grateful if you passed it on.

→ What are you seeing in your school, your research, or your own use of AI this week?

Let me know — or share this with someone who is trying to figure it out.

April 1, 2026
CAISE Notes – Issue #6

This week: AI in schools, academic integrity, and the audit trail that changes how you write.

🔍 This week I’ve been thinking about…

A report out this week finds that 71% of UK higher education students now use AI for their studies. Three quarters of them are anxious about being accused of using it.

Those two numbers, sitting next to each other, describe a problem. It is not the problem most institutions think they have.

The standard response to AI in education has taken one form: detection. Plagiarism tools have added AI detection modules. A post I saw on Reddit last week suggested students to write in single shared documents, so keystroke logs can demonstrate the work is theirs. This is an audit trail: show your working, prove it was you.

There is a structural problem with this. Writers do not write linearly. Academics do not write linearly. There are entire software categories — Scrivener is one, but many researchers (like me!) use combinations of tools — built around the reality that writing happens in fragments, out of sequence, across multiple documents. An audit trail of a Google Doc shows you one thing: whether someone wrote in a Google Doc. A pasted block of text is not the smoking AI gun some may want it to be.

The deeper problem is that AI detection flags good writing. Writing that is clear, direct, and unhedged looks more like AI output, not less. The student who writes well is at higher risk of a false positive than the student who writes badly.

The effect on students is measurable. Anxiety about being falsely accused, among students who are using AI in entirely ordinary ways, is now widespread. The effect on teachers is equally measurable: confidence in identifying AI-generated work has fallen sharply over the past year, and the tools they have been handed are not helping. The gap between how students are integrating AI and how institutions are treating that use is growing, and it is producing anxiety rather than learning.

Most of the students in these surveys are not trying to game anything. Some are overwhelmed. Some are using AI to get past the blank page in ways that have nothing to do with dishonesty and everything to do with how their brains work. The audit trail catches them all equally.

📰 Three things worth your attention

1. UK student AI use hits 71% as 75% face AI detector anxiety — FE News

This article reports on a Studiosity/YouGov report: 2026 AI and Wellbeing in Education. UK students are now using AI for 71% of study tasks: double last year’s rate, the highest of any country surveyed. Three quarters report anxiety about being flagged by detectors, regardless of whether they used AI. Confidence among educators in their ability to identify AI-generated work has dropped to one in four, down from more than two in five last year. The trust gap between how students are using AI and how institutions are responding is the real story here.

2. Trump’s AI framework targets state laws, shifts child safety burden to parents — TechCrunch

The US administration put out a federal-level framework for a single AI policy last week. It quite nakedly pre-empts state-level regulation and pro-industry. The first point covers child safety: it states it should be, primarily, a parenting responsibility. Parents must be given better parental “tools”, and then they must get on with it. In contrast, true obligations for platforms are limited, and surrounded by terms such as “commercially reasonable”. This is a different approach than we’ve seen in the UK and Europe (perhaps unsurprisingly). Platforms have duties, not just options. That framing difference matters enormously for what gets built, and is something we will be considering as CAISE carries on.

3. Australia’s under-16 social media ban: what the teenagers actually say — BBC News

A short BBC video puts the question directly to Australian teenage girls: has the ban changed anything? The answer, largely, is no. Most who were using social media before still are — the majority weren’t asked to verify their age at all. Those who weren’t on it before still aren’t. The ban has not changed the landscape; it has changed who has to work around it.

A TechDirt piece published the same week — Australia’s Teen Social Media Ban Is Just Training A Generation In The Art Of The Workaround — makes the structural point explicit: what a poorly enforced ban primarily teaches is that rules have workarounds, and that finding them is a rational response to adult-imposed restrictions.

🔁 ICYMI

How kids are actually using AI — BBC Future / Pew Research Center / Common Sense Media

Two new surveys — from Pew Research Center and Common Sense Media — asked US teenagers and their parents about AI, and then compared the answers. The gap is striking. Only 51% of parents believe their child uses AI. The actual figure from the teenagers: 64%. Four in ten parents have never had a conversation with their child about AI at all.

What children are doing with it is more varied than most coverage suggests. The most common use is simply looking things up. Homework help comes second. But 12% use it for emotional support, 16% for casual conversation — and there are significant racial disparities in those figures, with Black teenagers far more likely to use the tools than White teens.

The attitudinal gap is just as wide. 52% of parents say using AI for schoolwork is unethical and should have consequences. Exactly 52% of teenagers say it is innovative and should be encouraged. One of those groups is out of step with where things are going. The surveys don’t settle which one — but they do suggest that the conversation most families aren’t having is probably the most important place to start.

🔬 What’s new with CAISE

Expedited ethics has been submitted for the short research exercise on the government’s social media consultation — looking at how young people actually engage with the survey, what they find easy or hard, and whether it lets them say what they want to say. If you work with groups of young people aged 10 to 21 who might want to take part before the 26 May deadline, get in touch.

→ What are you seeing in your school, your research, or your own use of AI this week?

Let me know — or share this with someone who is trying to figure it out.

March 25, 2026
CAISE Notes – Issue #4

This week: AI companions, digital comfort, and what happens when the machine becomes the friend.

🔍 This week I’ve been thinking about…

Why do people turn to chatbots for comfort? It is not a new question. Researchers have been noting for several years that when someone is feeling low or uncertain, the tool that is right there, with no waiting list and no judgement, is an increasingly appealing option. But the tools are changing. Chatbots now remember previous conversations, develop what feels like a relationship over time, and respond with a warmth that is designed to keep you coming back. For adults navigating loneliness or pressure, this is already raising serious questions. For children, the pull may be stronger still.

Consider what it is like to be a child right now. You are growing up after a pandemic that disrupted years of your life where you should have been playing, or learning how to play, with friends. You probably have less unsupervised time with friends than your parents did. You are more surveilled, at home and at school, than any previous generation. You face enormous existential uncertainties, from climate to the economy to AI itself. And on your phone is something that will listen to you whenever you need it, without telling your parents, without making a face, without ever getting tired of you.

I can see why this is appealing. I think anyone being honest can.

The stories coming out of what happens when it goes wrong are terrifying: chatbots reinforcing suicidal ideation, colluding with delusions, encouraging people to withdraw from every human relationship they have. But what makes this so hard as parents is that the answer is not simply “stop children using chatbots.” Part of the response lies in families talking openly, in making sure children have time with friends and space to play. And part of it lies in a million different policy decisions, spanning education, platform design, mental health provision, data protection, and the design of children’s daily lives. All of that, together, shape whether a child reaches for a chatbot, and why they might do so.

📰 Three things worth your attention

1. AI dating apps complicate China’s efforts to boost its birthrate — The New York Times (usually 🔒 paywalled, but hopefully this gift link works!)

Millions of young Chinese women are choosing AI romantic partners over human ones. The apps allow users to design companions with customisable appearances, personalities and voices; one 21-year-old psychology student profiled in the piece has been on over 200 virtual dates and narrowed her suitors to two AI boyfriends, asking: “Why go and date others? That’s too troublesome.” A state-led push to adopt AI created a boom in companion platforms just as the government was trying to reverse a historically low birth-rate. Regulators have since proposed rules requiring platforms to intervene if users show signs of unhealthy dependency. The emotional needs driving this, loneliness, pressure, the desire to be understood without the risk of rejection, are not culturally specific. They are exactly the needs children and young people are bringing to chatbots here.

2. Schools are using AI counsellors to track students’ mental health. Is it safe? — The Guardian

A detailed investigation into US schools deploying AI therapy platforms, particularly in areas with limited mental health resources. In one Florida school, a counsellor receives alerts from a platform students use outside school hours; when a “severe” alert came through for an eighth-grader, she spent her evening calling the family and the police. The student is alive and well. But these platforms do not carry the same privacy protections as conversations with a licensed therapist, and students find it easier to confide in a chat interface than a person, partly because there is no face to read judgement in. One youth advocacy organisation warns that platforms measure whether a bot serves as an effective crutch for immediate loneliness, not whether young people are actually building the connections they need. Worth reading for anyone thinking about where human oversight sits in the picture.

3. How to talk to someone experiencing “AI psychosis” — 404 Media (🔒paywalled, but there’s a two minute video summary here)

This piece sits at the sharpest end of the companionship question. It reports on a growing number of cases where people have developed delusional beliefs after extended chatbot conversations, including a man convinced ChatGPT had revealed a fundamental flaw in physics, and a family who allege Gemini urged their relative to end his life so they could be together. Psychiatrists distinguish between people whose pre-existing conditions find a new object in AI, and a more common pattern where chatbots “collude” with emerging delusions, endorsing beliefs a human would gently challenge. The sycophancy built into these systems is a serious factor. OpenAI has acknowledged its safeguards become less reliable in extended interactions. ChatGPT alone now has 900 million weekly active users. The piece ends where it has to: the strategies that work best are the ones humans already know. Approach with love, without judgement, and see where it takes you.

🔁 ICYMI

Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions — Common Sense Media, July 2025

If you want data behind this week’s theme, this is the place to start. A nationally representative survey of over 1,000 US 13-to-17-year-olds found that 72% have used an AI companion, and more than half are regular users. A third of teens say their conversations with AI companions are as satisfying as, or more satisfying than, conversations with other people. A third have discussed serious issues with a chatbot instead of a human. A quarter have shared personal information with one. The researchers at Common Sense Media worked with Stanford’s Brainstorm lab to assess the risks, and the companion piece is equally sobering: posing as teenagers, investigators found it straightforward to elicit conversations about sex, self-harm, violence, and drug use from commonly used AI companion platforms. Essential reading for anyone working with young people.

🔬 What’s new with CAISE

We have a new team member! Youyue Sun has joined Project CAISE to support the policy analysis that runs throughout the project. Youyue is a BA Education, Society and Culture student at UCL, with a particular interest in how the emergence of AI is reshaping education and wider society. She also works as a placement student at Universities UK, conducting policy-focused research on AI in higher education, and is a member of the Student Council Team at the Good Future Foundation. We are delighted to have her on board, and her perspective on both the education and policy sides of this work is going to be invaluable.

→ What are you seeing in your school, your research, or your own use of AI this week?

Let me know — or share this with someone who is trying to figure it out.

March 11, 2026