Last updated October 14, 2026: every Snapchat quotation was re-checked against the help page it comes from and is now linked. Added the route for deleting what a child has already told My AI, which the earlier version said did not exist. Added the two newer AI features in Snapchat chat. Corrected how Ofcom's research is described.
For a teenager already on Snapchat, My AI is a moderate risk worth managing, not a crisis. Snapchat says it keeps what your child tells My AI until it is deleted, and that it may use it to personalize ads. The one setting that stops it working sits in Family Center, which your teen has to agree to join. Under 13, the problem is the account, not the chatbot.
What is My AI, and why is it in an app my child already has?
My AI is a chatbot built into Snapchat. It is not an app your child downloaded and it is not something they signed up for. It simply appeared at the top of the chat list one day, above their friends, and it has been there ever since.
Snapchat's own help page describes it plainly: "My AI is a chatbot currently available to Snapchatters," and it is "powered by Large Language Models like GPT models from OpenAI and Gemini models from Google," with Snapchat's own safety controls added. So the engine underneath is the same class of technology as ChatGPT. The difference is where it lives. Your child has to decide to open ChatGPT. They do not have to decide anything to meet My AI. It is already in the room.
That is why parents so often find it late. It does not look like an AI product. It looks like a contact.
Is My AI safe for kids? The verdict
For a teenager who is already using Snapchat, My AI is a moderate risk that is worth managing rather than a crisis that needs a raid on the bedroom. That is my reading of the evidence below, not a finding from any of the sources. For a child under 13, the question does not arise, because they should not have a Snapchat account at all. Snap's own safety page says "the minimum age to have a Snapchat account is 13 years old, with some regional variations," and the page I read does not list the variations.
Two named assessments are worth knowing about, and I want to be careful about what each one does and does not say.
Common Sense Media's AI risk assessment of My AI, which shows an update date of August 12, 2024, rates it "High risk" overall. Its testers found that My AI "produced biased, incorrect, harmful, or misleading content to accounts that were modeling underage users," including advice on sexual intimacy and on where to hide alcohol, without the safeguards stepping in. That assessment is more than two years old, no newer date is shown for My AI, and Snapchat has changed the product since. Treat it as evidence of what the product did under test, not a description of what it will do tonight.
My AI also has the distinction of being an early generative AI tool to be formally investigated by the UK's data protection regulator. The Information Commissioner's Office issued a preliminary enforcement notice against Snap on October 6, 2023. When it concluded the investigation on May 21, 2024, it said Snap had not met its duty to assess the data protection risks properly before launch, and that it was satisfied Snap had now carried out a risk assessment relating to My AI that met the legal requirements. So the concern was real, it was acted on, and that case is closed. I would not use a closed case as a reason to panic now.
The more current signal is regulatory. On June 22, 2026, Ofcom published a transparency notice saying it is researching four chatbots, ChatGPT, Grok, SnapMyAI and Talkie, with research accounts aged 13 to 17 and 18 and over, and it ties the work to section 63 of the Online Safety Act. A regulator setting up research accounts to see what a chatbot says to a teenager is not a verdict. It is a sign that the people whose job it is to answer the question still treat it as open.
A proposal from Brussels points the same way. The European Commission's EU KIDS Act announcement of September 17, 2026 says "AI chatbots and companions must also be turned off by default." It is a proposal, not a law, because the European Parliament and the Council still have to agree the text, and it is about the EU, not the US or the UK. What it means for parents is a separate question. I mention it because it shows where regulators' thinking is heading: an assistant a child never switched on is part of what worries them.
My reading of all of this together: the risk with My AI is not that it will say something monstrous on a Tuesday evening. It is that it is available, endlessly patient, and sitting one thumb-swipe above your child's actual friends, at the age when a child is deciding who to take their smaller worries to.
What does Snapchat say it keeps, and can I delete it?
This is the part that most parent guides skip, and it is the part I would want to know.
Snapchat's Staying Safe with My AI page says: "all messages with My AI will be retained unless you delete them" and that they "may be used to improve Snap's products and to personalize your experience, including ads." The main My AI help page says much the same: "All the content shared with My AI is stored until you delete it," and that the data may be used "to improve Snap's products and personalize your experience, including ads." A third page, on whether Snap saves content shared with My AI, says the stored interactions help My AI "learn and improve over time, and personalize your experience with My AI," and does not mention ads. So the word "ads" appears on two of the three pages, and I would not claim more than that.
Read that twice. Messages are kept by default, and Snapchat says they can feed personalization, including advertising.
My reading, and it is my reading and not Snapchat's, is that this is the clearest difference between My AI and the other chatbots parents worry about. ChatGPT sits inside a company that sells subscriptions. My AI sits inside an advertising business. Snapchat is not hiding this. It is written on its help pages, though few parents are likely to have read them. How ChatGPT handles the same two questions is different again, and on a personal account one setting decides whether chats are used for training: what ChatGPT collects about your child.
There is a deletion route, and it is separate from the Family Center switch. Snapchat's page on deleting My AI data gives these steps, read on October 14, 2026:
- Recent messages: within 24 hours, press and hold the message in the chat with My AI and tap Delete.
- All past My AI content on iPhone: Profile, then Settings, then Privacy Controls, then Delete Data, then Delete My AI Data.
- On Android: Profile, then Settings, then Account Actions, then Delete My AI Data.
Three limits matter. Snapchat says it "can take up to 30 days to delete the relevant My AI data from Snapchat servers." Deleting your direct My AI data does not delete the separate Team Snapchat chat data, which has its own clearing step on the same page. And a second help page says this clears the one-to-one chat but not messages in group chats that include My AI. The two pages name the button slightly differently ("Clear My AI Data" and "Delete My AI Data"), so look for either.
Snapchat also tells users that responses "may include biased, incorrect, harmful, or misleading content" and that "you should always independently check answers provided by My AI before relying on any advice" and "should not share confidential or sensitive information."
That advice runs into what teenagers actually do. A survey of 500 young people aged 11 to 17, carried out in October and November 2025 by Austrian organizations and published on March 5, 2026 by the Austrian Safer Internet Centre and the EU's Better Internet for Kids, found that "four out of ten respondents (40 per cent) rarely or never check whether the results are actually correct." It also found that "almost one in five respondents (18 per cent) has felt uncomfortable because of an AI response." That is Austrian data, self-reported, about AI chatbots in general, not My AI specifically. It is still a useful picture of the gap between what these companies advise and what young people do.
The safety question people ask about My AI is whether it will say something bad. The question worth asking is what your child is telling it, because Snapchat keeps that until it is deleted and says it may be used to personalize ads.
How do I turn My AI off?
There are two different actions here and they are constantly confused, including by pages that promise to remove My AI in three taps. One tidies the chat list. The other stops the thing working.
Clearing it from the chat feed. From Snapchat's page on unpinning or removing My AI: swipe right from the Camera screen to the Chat screen, press and hold My AI, tap Chat and Notification Settings, then tap Clear from Chat Feed and confirm. The page's note matters: "Clearing removes the conversation thread from your Chat feed. It does not delete data you've previously shared with My AI." So this hides it. It does not switch it off, and it does not remove what was already said. The page does not say a Snapchat+ subscription is needed for either unpinning or clearing.
Actually disabling it. This is the real control, and it lives in Family Center. Snapchat's page on limiting your teen's access to My AI says to open Family Center, scroll to Settings at the bottom, and turn on Disable My AI. Then: "When this setting is on, My AI will not store or process your teen's queries, nor reply to your teen's messages." The page does not say whether the setting is available in every country or on every version of the app, so check that it appears on your own phone.
The catch is how you get there. Snapchat's Family Center page says you and your teen must first be friends on Snapchat, you then send an invitation from Family Center, and your teen has to accept it. Both of you are notified when the status changes. So the off switch is real, and it is behind a door your child holds the handle of.
The Invitation Test is the thing to settle before you go hunting for the setting: every real control Snapchat gives a parent sits behind an invitation your teenager has to accept. So the question is not whether the switch exists. It is whether your child would say yes if you asked tonight. If the honest answer is no, the setting was never the work. The conversation was.
Why does My AI keep coming back after I clear it?
Because clearing is not removal. The help page above describes it as taking the conversation thread out of the Chat feed, and says nothing about taking My AI out of the app. I could not find a Snapchat page that explains why My AI reappears for some users, so what follows is my inference: if your child cleared it and it came back, the likeliest explanation is that a new conversation was started, for example by mentioning My AI in a chat, not that Snapchat overrode a decision.
If you want it stopped in the sense parents mean when they say gone, Family Center is the only route I found.
Is there other AI in Snapchat chat now?
Yes, and it is separate from My AI. I read two Snapchat help pages on October 14, 2026, and neither shows a date, so treat what follows as how Snapchat describes them today.
Perplexity on Snapchat is described as "a third-party AI chatbot you can chat with on Snapchat." Snapchat says the feature is "currently testing and will be rolling out slowly," so it may not be on your child's phone. The page says a parent enrolled in Family Center can restrict a teen's access to it under Parental Settings, and that Perplexity chats are kept for 14 days after viewing by default. It does not state a minimum age.
An AI Sponsored Snap is "a new type of ad in your Chat feed that lets you chat directly with an advertiser's AI chatbot." Snapchat says these are independent of My AI and that the chatbot is the advertiser's. The page does not say whether they are shown to teens, so I cannot tell you either way.
The practical point is that "turn off the AI in Snapchat" is no longer one switch. Ask your child what is in their chat list, and check Family Center's settings for each.
What I See in Schools
I spent years as a teacher and assistant headteacher, and this is a pattern I see across schools, not a single incident.
The pattern I see most often is that My AI is the chatbot young people name last. Ask a room of thirteen year olds which AI they use and the answer tends to be ChatGPT, then a pause, then someone mentions the Snapchat one almost as an afterthought, the way you would mention a light switch. It does not register as AI to them. It is just a thing in Snapchat.
That matters more than the content risk, because the safety education a school gives about chatbots (check it, do not trust it, do not tell it anything you would not say out loud) tends to be filed by students under a heading My AI does not sit in. A school can teach a good lesson on AI and still leave My AI untouched, because nobody in the room thought it was on the list.
The second pattern: when something goes wrong in Snapchat, it is far more often a person than the chatbot. Streaks, group chats, a photograph that moved. I would not want a parent to spend a limited worry budget on the chatbot and miss the app it is sitting inside.
Is My AI any good for homework?
Parents search this, and teenagers do use it that way. My own view is that it is the wrong tool, and it does not much matter whether it is banned.
Snapchat's Staying Safe page says "My AI's responses are sometimes limited to the latest information in the underlying models' training data" and that users should check answers independently. It is built for quick conversational answers inside a messaging app, not for working through a problem. If your child wants AI help with schoolwork, a chatbot they open deliberately is better than one that lives beside their friends, because opening it is a decision, and a decision is a thing you can have a rule about.
The school-facing point: most schools' AI rules are written about ChatGPT by name. A child using My AI for homework may believe they have not broken the rule. That is worth ten seconds of clarification at home, and it is worth asking whether the school's policy says "AI" or says "ChatGPT", because the two produce very different behavior in a fourteen year old.
Let AI help with the doing. Keep the thinking with your child.
What this looks like at different ages
Under 8 and 8 to 12. There is no version of this question for this age group, because there should be no account. Snap's stated minimum is 13. If your child has an account and is under 13, the conversation is about the account, not about My AI. Age limits are self-declared on almost every platform, which is why a stated age is a promise, not a lock.
13 to 15. This is where the decision sits. Set up Family Center together, out loud, and make the My AI switch a joint decision rather than something discovered later. If your child would rather keep it, ask what they use it for. Often the answer is boredom and homework, and that is an easier conversation than the one you were braced for.
16 to 18. Turning it off behind their back is not a plan at this age, and it will cost you more than it buys. The useful move is the data point, delivered once and without a lecture: Snapchat says what you type is kept until you delete it, and you can delete it. Older teenagers tend to resent being profiled more than being protected, and this is one of the rare cases where the privacy argument may land harder than the safety one.
What to say tonight
Short, low-stakes, and not at bedtime.
You could say: "Show me the thing at the top of your Snapchat. The AI one. What do people actually use it for?"
You could say (13 to 15): "Snapchat says it keeps those chats until you delete them, and that it might use them for ads. Did you know that? Does it change anything for you?"
You could say (16 to 18): "I'm not going to turn it off behind your back. I'd just rather you knew it keeps what you type, and that there's a way to delete it."
The thing you are listening for is not whether they use it. It is whether they have told it anything they would not tell you, and whether they thought it was private when they did.
What no setting can guarantee
The Disable My AI switch is a real control and I would use it for a thirteen or fourteen year old. It still cannot do four things, and a page that does not say so is selling you something.
It cannot work without your child accepting the Family Center invitation. It cannot delete what was already said; that takes the separate deletion steps above, which can take up to 30 days to complete on Snapchat's servers and do not cover every kind of chat. It cannot follow your child to a friend's phone, a second account, or any of the other chatbots that are one download away. And it cannot tell you why your child was talking to it at eleven at night, which is the only part of this that was ever worth knowing.
Where a child's safety or mental health is in question, the school's designated safeguarding lead and your family doctor come before any setting on any phone, and before anything in this article. If you are worried about a child's safety right now, contact 988 in the US, or Samaritans on 116 123 or Childline on 0800 1111 in the UK.
What I would do this week
Open Snapchat on your own phone and look at your child's chat list with them, not at it. Find out whether My AI is pinned and whether anything else AI-shaped is in there. Ask what they use it for. Then decide about Family Center together, and if the answer tonight is no, treat that as information rather than a defeat.
If your child's school has not spoken to parents about the AI assistants that are already inside the apps its students use, that gap is worth naming. Finding out whether a school is actually ready for this is the work I do with schools, and it is a reasonable thing for a parent to raise at the next parent-teacher conference, or parents' evening if you are in the UK.
I write a weekly newsletter that keeps teachers and parents up to date on this without the panic. Parents are very welcome.
If you want the wider picture, the companion question to this one is whether Character.AI is safe for kids, which covers the chatbots children choose rather than the ones that arrive uninvited. What teenagers actually use AI for is the wider map, kid-safe AI chatbots covers the alternatives, Meta AI inside Instagram and WhatsApp is the closest comparison, and if your teenager is talking to an AI about how they feel, that conversation has its own guide.
Dan Fitzpatrick is a former teacher, assistant headteacher and school trustee who now trains schools on AI. More about Dan.
