The First Disclosure May Be to a Machine

Guest written by Wolfpack member, Tania Mackie

‍AI literacy is safeguarding now.

A child's first disclosure can now be typed into a chatbot, not spoken to a teacher, not shared with a parent, not written on a note passed quietly across a classroom, but typed into a prompt box late at night, when the child is frightened, curious, lonely or simply unsure who else to ask, and that changes safeguarding now.

The conversation has already started

Artificial intelligence is not waiting outside the school gates while adults decide whether it belongs in education, because it is already in bedrooms, browsers, search results, homework and conversations. Ofcom's 2026 research found that 56 per cent of children aged 8 to 17 have used AI, and among those, 11 per cent have used it as someone to talk to or as a friend, which is not simply a usage statistic but a safeguarding signal. Children are using these tools to learn, create and explore, and they are also using them to test questions, seek reassurance and make sense of things they may not yet be ready to say aloud, so while AI may be where a child first forms the question, a trusted adult must still be where they find human care, context and help.

The system is looking in the wrong place

Schools have filed technology, curriculum and safeguarding in separate cabinets, and yet the child carries all three in the same pocket, so a conversation with an AI tool can move from homework to friendship, body image, anxiety or self-harm in seconds, and the technology does not pause while we work out which policy owns the issue. The problem is not teachers, who are already absorbing another enormous change while holding relationships, learning and wellbeing together every day, but a system that still treats AI as a technology project, and that system is not broken so much as functioning exactly as it was designed to, built for an era that has already gone, which makes it not broken but irrelevant. This is a leadership issue, a curriculum issue and a safeguarding issue, and all three must finally meet.

We may be asking the wrong question

We spend a great deal of time debating whether young people should use AI, when the more important question is whether we are teaching them to use it well, because as educators we have spent decades developing children's ability to question, analyse, evaluate evidence, solve problems, create and reflect, and those skills have never mattered more. AI does not replace Bloom's Taxonomy, it makes the higher order thinking within it more valuable than ever, because a child who knows how to open an AI tool is not necessarily equipped to question its answer, and confidence is not competence, just as access is not literacy. Ofcom found that 46 per cent of children aged 8 to 17 who had read AI summaries in search results believed those summaries were always accurate, yet only 23 per cent recalled being taught how to spot AI generated content, and that gap is where the harm begins. A machine can produce an answer in seconds, but the learner must still decide whether it is accurate, biased, incomplete or entirely wrong, which is analysis, then compare sources, test claims and consider what evidence is missing, which is evaluation, and finally take what is useful, reject what is not and build something new, which is creation, so Bloom's Taxonomy was built for exactly this moment.

Good prompting is clear thinking

I now hold different conversations with different AI tools, giving them different levels of context, asking different questions and challenging their answers in different ways, and the tools have not simply changed how I work, they have made me more conscious of how I think. Good prompting is not magic, it is communication, analysis and refinement, knowing what information matters, what question to ask and when an answer is not good enough, and if experienced professionals are still learning how to get the best from these tools, we should not underestimate how overwhelming it can be for a young person, which is why our children deserve to learn how to think with AI rather than simply how to operate it.

Trusted adults matter more, not less

Technology can identify patterns, surface concerns and help adults notice what might otherwise remain hidden, but it cannot know the whole child, and it cannot understand the look exchanged at the classroom door, the sudden withdrawal from friends or the reason a piece of writing feels different today, because safeguarding has never relied on a single signal, but on professional curiosity, context and adults who notice. England's DfE product safety standards recognise this, expecting learner facing AI products to identify possible safeguarding disclosures, alert appropriate staff and direct distressed learners towards human help, so the technology should strengthen the pathway to a trusted adult, and it must never become that pathway's final destination.

Five moves we can make now

1. Put AI inside safeguarding, so that tools, prompts, disclosures and patterns of use sit in the policy, the training and the governance conversation.
2. Teach the thinking rather than the buttons, using Bloom's Taxonomy to develop questioning, analysis, evaluation, creativity and reflection.
3. Make prompting visible, modelling how we add context, refine questions, challenge answers and reject the weak ones.
4. Rehearse the route to help, so that every child knows which trusted adult to approach and what will happen when they share a concern.
5. Keep the monitoring human, letting technology help us notice, and then applying professional judgement, context and care.

This is the work. The choice is not between protecting young people and preparing them for a world shaped by AI, because we must do both, which means we need safeguards around the technology, skilled adults around the child and an education that develops curiosity without ever surrendering judgement. So yes, let us teach young people about AI, and more importantly, let us teach them how to think, how to question and how to reach for another human being when the answer on the screen is not enough, because the first disclosure may be to a machine, but what happens next must be profoundly human, and AI literacy is safeguarding now.