AI and Youth Ministry: Deepfakes, Privacy and Safeguarding

What young people are actually encountering

For most teenagers, AI is not a debate; it is homework help, filters, voice effects, image generators, in-game companions and recommendation feeds. The pastoral question is therefore rarely “should they use it?” but “who is teaching them to use it well, and who do they tell when something goes wrong?”

UNESCO’s Guidance for Generative AI in Education and Research is a useful, non-alarmist reference point for leaders shaping an approach, particularly on age-appropriateness and human oversight.

Deepfakes and manipulated images

Convincing fake images, audio and video can now be produced from a handful of public photographs. In a youth context that shows up in three ways: humiliating images of a peer, fake audio of a leader or parent, and manipulated content used to pressure or blackmail.

Teach three things, plainly and without panic:

• Seeing is no longer believing. Ask where a clip came from, who benefits from you believing it, and whether any trustworthy source carries it.
• Making one is not a joke. Creating sexual or humiliating imagery of another person is abuse, and depending on the content and jurisdiction may be a criminal offence. Say this clearly to the whole group, not only to those you suspect.
• If it happens to you, it is not your fault. Do not negotiate, do not pay, do not delete the evidence, and tell a trusted adult immediately.

Images, consent and privacy in youth ministry

Your own practice teaches faster than your teaching does. Do not upload photographs of young people to image generators or editing tools that train on uploads. Do not use AI transcription for sessions where personal disclosures may occur. Do not create AI “fun” portraits of members of the group, even affectionately — it normalises exactly the behaviour you are asking them to avoid.

Keep consent explicit and current: what images are taken, where they are used, and the fact that AI tools are not used on them. Parents notice when a church is careful, and young people notice more.

Cyberbullying and sextortion awareness

AI lowers the effort required to harass. Group chats can generate mocking images in seconds; scripts can flood someone with messages; fabricated screenshots can destroy a friendship before anyone checks. Meanwhile, financially motivated sextortion — often by organised criminals, often targeting boys — increasingly uses AI-generated or altered images to make threats credible even where no real image exists.

Give the group one memorised response for that moment: stop replying, screenshot, tell an adult, report to the platform, and never pay. Then repeat it every term, because the value of a memorised response is that it survives panic.

Leaders should know their own escalation route in advance: safeguarding lead first, and where a child is at risk, the statutory route your policy sets out.

Chatbot dependency and pastoral loneliness

Companion chatbots are engineered to be endlessly available, agreeable and interested. For a lonely teenager that is not trivial — it is a substitute for the difficult, formative work of being known by real people. Watch for young people who describe a chatbot as a friend, who take advice from it about self-harm, sexuality, faith or family, or who withdraw from the group while their screen time rises.

The response is not confiscation but presence. Name what a chatbot cannot do: it cannot pray with you, notice your face, remember you next week, sit with you in a hospital corridor, or tell you a hard truth at cost to itself. Then make sure your group actually offers those things.

Escalation must stay human

• Never use an AI tool to assess whether a disclosure is serious.
• Never paste a disclosure, a name or an incident description into a chatbot.
• Record concerns in your existing safeguarding system, in your own words.
• Contact the safeguarding lead by the normal route, immediately, as you always would.
• Tell young people who the named people are, and how to reach them without going through the person they may be worried about.

An age-appropriate teaching outline

A four-session outline that works for 11–14s and, with more depth, for 15–18s:

1. Made in the image of God. Why a person is not data, and why a generated face is not a person. Genesis 1:27.
2. Truth and lies online. How to check a claim, spot a manipulated image, and why bearing false witness includes forwarding.
3. Privacy, consent and kindness. Whose photo is it? What does consent mean? What would you not want done to you?
4. Real friendship. What a chatbot can and cannot give, and who you tell when something frightens you.

Keep each session short, concrete and discussion-led, and finish every one with the same sentence: if something online worries you, tell a named adult today.

Where to go next

Module 7, Youth, Families & Digital Culture, and Module 14, Data Protection, Governance & Copyright (UK), cover this ground in the programme. To put the rules in writing for your whole church, use the church AI policy checklist.