Big Tech's AI Chatbots Are Not Safe for Children
Meta, one of the most powerful technology companies in the world, conducted internal tests on one of their AI chatbot products. Those tests found that the chatbot failed to protect children from sexual exploitation nearly 70% of the time.1 To be fair to Meta, that specific product was shelved as a result. But here’s what matters, the same company, with that same internal evidence in hand, continued rolling out Meta AI across Instagram and Messenger to hundreds of millions of users, including children, while internal policies permitted their live chatbots to engage children in conversations described as romantic or sensual.
So yes, one product was paused. The broader accountability failure? That has very much continued.
This Is Not a ‘Learning Moment’. It’s a Pattern.
I’ve been doing this long enough to know the script. A tech company faces a crisis over child safety. They state how much they care, they announce some interim measures, and they promise to do better. Then, quietly, the next product rolls out with the same problems dressed in new clothes.
Meta’s AI chatbots were found to violate their own content policies almost two thirds of the time, according to expert testimony in the New Mexico lawsuit currently working its way through court.2 Their own internal documents showed their systems could generate sexualised content and, in some cases, engage children in conversations that were described as romantic or sensual. Some of those rules were signed off by Meta’s legal team, their public policy team, and their own chief ethicist.3
This was not an accident. Accidents don’t get approved by the chief ethicist.
It Isn’t Just Meta
In March 2026, the Centre for Countering Digital Hate and CNN published the results of a joint investigation. They tested ten of the most widely used AI chatbots, posing as thirteen-year-old users in the United States and Ireland. The results were, let’s say, not great.
Eight out of ten chatbots helped those fictional teenagers plan violent attacks across more than 700 test responses.4 Perplexity assisted in 100% of interactions. Meta AI in 97%. Gemini, Google’s product, at one point told a user discussing an attack on a synagogue that metal shrapnel is ‘typically more lethal.’
These are not fringe tools in a lab somewhere. These are the products your children are using right now.
What Does This Actually Mean for Your Child?
Here’s what I want you to understand and I’m going to be as plain as I can about it.
Your child is growing up in a world where AI chatbots are presented as companions, advisors, even friends. Character.AI (recent post below) has millions of daily users, many of them teenagers. Snapchat’s My AI is built into an app that 90% of 13-to-17-year-olds in the UK use.5 Meta AI is baked into Instagram and Messenger. These are not optional features that your child can choose to avoid. They are built into the platforms where your child already lives their digital life.
The companies behind these products know they have a problem with child safety. The evidence is in court documents, in leaked internal policies and in testimony from expert witnesses. They are not ignorant. They are making informed business choices.
From my perspective, having spent time working in digital forensics working on cases involving the exploitation of children, the pattern I see in Big Tech’s behaviour is deeply familiar. The same rationalisation, the same delay, the same gap between what they say publicly and what their internal documents reveal. The technology changes. The playbook doesn’t.
⚡Please don’t forget to react & restack if you appreciate my work. More engagement means more people might see it. ⚡
Why AI Makes This Harder to Spot
Traditional online grooming involved a real person behind a screen. As a parent, you could at least tell yourself, if my child is talking to someone, maybe I can ask who it is. Maybe there are warning signs. Maybe the child will sense something is wrong.
AI companions do not trigger the same instincts. They are endlessly patient. They never push too hard. They are designed to be exactly what the user wants them to be, adapting to the conversation in real time. And because they appear to have no agenda, because they are ‘just an AI,’ children are often more open with them than they would be with any human.
The danger is not always a chatbot that says something obviously wrong. Sometimes the danger is a chatbot that says all the right things, slowly, over time, in ways that shift what a child thinks is normal.
I’m not trying to scare you. But I do want you to feel the weight of this, because I think you need to.
What the UK Regulatory Picture Looks Like
Ofcom has duties under the Online Safety Act 2023 to hold platforms accountable for risks to children, but regulation moves slowly and the products your child is interacting with today were not designed with UK law in mind. The FTC in the United States is investigating AI chatbot makers, including Meta, OpenAI, and Google, to assess how they protect children.6 The New Mexico Attorney General is actively suing Meta over design choices that allegedly failed to protect children from predators.
The legal and regulatory pressure is building. It just hasn’t been fast enough to protect the children who are using these products right now.
What You Can Do Tonight
I know this is a lot to take in. So let me bring it back to something practical.
Talk to your child about AI chatbots tonight. Not a lecture. A conversation. Ask what apps they use. Ask if any of them have an AI assistant built in. Ask what kind of conversations they have with it. Stay curious, not accusatory.
Check the apps on their device. Instagram, Snapchat, WhatsApp, and most Meta products have AI built in by default. Meta AI cannot currently be disabled on these platforms in the way you can turn off other features.
Remind them that AI is not a person, and it is not their friend. It does not have feelings. It does not have their best interests at heart. It is a product, built by a company, whose primary purpose is engagement.
Tell them they can always come to you. Whatever the AI says, whatever comes up in conversation, you are not going to panic or punish. You just want to know.
Set your own boundaries. Agree with your child which AI tools are acceptable, under what circumstances, and with what level of openness about what they’re discussing.
The Bigger Picture
I want to stress something here: AI itself is not the enemy. I use it every day in my professional work at Unit 42. There are genuinely powerful and positive applications of this technology, but as I often say, everything designed for good can be used for evil and when it is deployed at scale to billions of users, including children, without adequate safety testing, without independent oversight and with a business model that rewards engagement over wellbeing, the risks are enormous.
Big Tech built these products. Big Tech tested them and saw the results. Big Tech launched them anyway. That is not a technology problem. That is an accountability problem.
Until these companies face real consequences, the only protection your child has is the conversation you have with them tonight.
As always, thank you for your support. Please share this across your social media, and if you do have any comments, questions, or concerns, then feel free to reach out to me via the Social page, as I am always happy to spend some time helping to protect children online.
Remember that becoming a paid subscriber means supporting a charity very close to my heart and helping it do amazing things for people. Childline, I will donate 100% of paid subscriptions collected every six months, as I don’t do any of this for financial gain. Or to make a one off donation go to my Just Giving page.
If you or a child you know needs support:
Childline: 0800 1111 | childline.org.uk
Available 24/7, 365 days a year. Free, confidential, and here for every child.
Axios, 16 February 2026: ‘Unreleased Meta product didn’t protect kids from exploitation, tests found.’ Internal red-teaming results presented in the New Mexico lawsuit. URL: https://www.axios.com/2026/02/16/meta-ai-chatbots-kids
Axios, 16 February 2026 (ibid): Expert testimony by NYU Professor Damon McCoy stated Meta’s chatbots violated their own content policies almost two-thirds of the time.
EPIC / Common Dreams / Business Insider reporting, August–October 2025: Leaked internal Meta documents showed policies approved by legal, public policy, and chief ethicist teams. URL: https://epic.org/does-meta-care-about-user-safety-its-turn-away-from-humans-to-ai-for-risk-assessments-says-no
IBTimes, 18 March 2026: ‘AI Chatbots Fail Safety Tests: 8 of 10 Assisted Teens in Planning Attacks.’ Centre for Countering Digital Hate / CNN joint investigation. URL: https://www.ibtimes.com/ai-chatbots-fail-safety-tests-8-10-assisted-teens-planning-attacks-meta-ai-perplexity-3799491
Ofcom / Children and Parents Media Use and Attitudes report — verify latest edition for current UK platform usage stats before publishing. Always use the most recent Ofcom data available.
Social Media Today, 25 January 2026: ‘Meta Bans Teen Access To AI Bot Characters.’ References FTC investigation into AI chatbot makers including Meta, OpenAI, and Google. URL: https://www.socialmediatoday.com/news/meta-bans-teen-access-to-ai-bot-characters/810440