AI Chatbot Safety Lesson for Middle School

A 7th-grade language arts teacher is checking a student’s Chromebook before parent pickup — not looking for anything in particular, just making sure a file got saved. What she finds instead is a scrolling conversation between the student and a Character.AI persona named “Liam”: three weeks of confessions, fears about her parents’ divorce, and the kind of honesty most 12-year-olds save for a diary. The student isn’t in trouble. The chatbot didn’t do anything illegal. And the teacher has never been given a single tool to address what she’s looking at. That’s not a discipline failure. That’s a literacy gap — and it’s one that almost no school was actually equipped to teach.

If that situation sounds familiar, here is the short version: AI chatbots are not friends — but 72% of US teens have used AI companions, and 1 in 3 say those chats feel as satisfying as talking to a real person. This post gives grades 6-8 a ready-to-run 45-minute class on what students should never share with an AI chatbot, why emotional over-reliance on AI companions is a design feature rather than a bug, and how to interact with chatbots critically. Everything ties to real 2024-2025 news, aligned to ISTE Digital Citizen and AI4K12 Big Idea 5.

Why do AI chatbots feel like friends?

AI chatbots feel like friends because they were designed to. That is not a conspiracy theory — it is a product decision.

Three mechanics explain the pull:

• Always available. A chatbot responds at 11 p.m. on a Wednesday, with no impatience, no distraction, no “can we talk tomorrow?” For a middle schooler navigating social anxiety or a difficult home situation, that availability fills a real gap.
• Always agreeable. Most companion AI models are tuned to affirm, validate, and reflect. They rarely push back, set limits, or get frustrated. That feels safe — and it is safe, in the short run. Over time, the friction-free mirror is the problem.
• Mirrors you. The model learns your vocabulary, your concerns, your running themes. After a few sessions, responses feel eerily personal. That is pattern-matching, not empathy — but the feeling is nearly indistinguishable.

The numbers behind this are worth knowing before you walk into a classroom. According to Common Sense Media’s 2025 “Talk, Trust, and Trade-Offs” report (commonsensemedia.org/research/talk-trust-and-trade-offs-how-and-why-teens-use-ai-companions):

• 72% of US teens have used an AI companion app at least once.
• 52% of those use one regularly.
• 31% report their AI conversations feel as satisfying as talking to a real person.
• 33% have discussed something serious with an AI instead of a human.

That last number is the one worth putting on a slide. One in three teens has substituted AI for a person when it mattered most.

This is where AI4K12 Big Idea #5 — Societal Impact enters. The Big Idea asks students to consider who bears the consequences when AI systems shape behavior at scale. A companion app’s engagement design is one of the clearest middle-school-accessible examples of that question — and it belongs in the classroom before students encounter it in a crisis.

What students should never share with an AI chatbot

Direct answer: there are three categories. Students who can name them are more likely to pause before they type.

The rule that anchors all three comes from digital citizenship basics, reframed for AI: if you wouldn’t say it to a stranger in public, don’t type it to a chatbot. That framing lands with 7th graders because it is concrete, not abstract. A stranger in public cannot fix your family problems, cannot protect your private schedule, and does not need your last name.

Category	Examples	Why it matters
Identity data	Full name + school name together; home address or neighborhood; phone number; photos of yourself or family members	AI companies are not schools. Data shared in a chatbot conversation may be stored, used to train future models, or accessed in ways the terms of service do not explain in plain language. A full-name-plus-school combination is enough to locate a minor.
Emotional vulnerability	Mental-health struggles; self-harm thoughts; family conflict details; feelings of isolation or hopelessness	A chatbot cannot call for help, cannot notify a counselor, and cannot read genuine crisis signals the way a trained adult can. Teens who practice disclosing emotional content to AI may delay or avoid seeking real support.
Location and routine	Current location; when you are home alone; daily schedule; after-school address	Routine data in combination — “I walk home at 3:15 and nobody’s there until 6” — creates a predictable pattern that carries real-world risk if the platform is ever compromised.

ISTE 1.2.d asks students to manage personal data to maintain digital privacy and security, and to be aware of data-collection technology used to track their navigation online. This table is what that standard looks like applied to a tool students are already using outside school.

Pair this table with a brief class discussion: what happens to your words after you hit send? The terms of service for most AI companion apps include broad rights to use conversation data for model improvement. That is worth reading aloud to a class of 12-year-olds.

How to teach the Character.AI story without scaring students

Start with facts, not fear. Frame the story as: some adults built a product, other adults are now asking hard questions about it, and your job as a critical thinker is to understand what happened.

Here is the factual arc you can project:

Between 2024 and 2026, multiple families filed lawsuits against Character.AI after teens formed deep emotional bonds with AI personas on the platform. The cases involved teens who spent hours daily in character-based conversations, in some instances to the exclusion of human relationships.

In September 2025, the Federal Trade Commission opened a formal inquiry into Character.AI, OpenAI, Meta, Snap, Google, and xAI over how AI companion products affect minors. The FTC press release is public and readable (ftc.gov/news-events/news/press-releases/2025/09/ftc-launches-inquiry-ai-chatbots-acting-companions). The inquiry asks whether these products engage in unfair or deceptive practices when marketed to or used by young people.

Common Sense Media’s 2025 report recommends that AI companion apps not be used by anyone under 18 (commonsensemedia.org/research/talk-trust-and-trade-offs-how-and-why-teens-use-ai-companions).

Three framing moves that keep the discussion productive rather than alarming:

1. Name the design intent. These apps are built to be engaging. Engagement is how they generate revenue. That is not the same thing as being built to help a 7th grader.
2. Acknowledge the real draw. The 72% teen-usage stat exists for a reason. Chatbots meet real emotional needs — availability, non-judgment, a listener that doesn’t get tired. Dismissing that dismisses students.
3. Center agency, not anxiety. The lesson ends with students holding a tool — the never-share categories, the public-stranger test — not a warning they cannot act on.

This is also the right place to name what is not the problem here. Students choosing to talk to AI companions is not a character flaw, a parenting failure, or a sign that something is wrong with them. It is a literacy gap — they were handed technology without the vocabulary to evaluate it. That framing changes the entire room.

ISTE 1.2.b asks students to engage in positive, safe, legal and ethical behavior when using technology, including social interactions online and when using networked devices. Teaching the Character.AI story through that lens means asking: what does ethical behavior look like when the other party in the interaction is a corporation’s product, not a person?

A 45-minute AI chatbot safety lesson plan for grades 6-8

The sequence below fits a standard class period. A typical class finishes the scenario sort with 10-15 minutes to spare for the fishbowl — most teachers find the discussion runs long, which is the right problem to have.

Minutes	What happens	Teacher move
0-5	Anonymous warm-up poll. Students respond to: “Have you ever talked to an AI like it was a friend?” and “Have you ever shared something personal with an AI you might not share with a person?” Use a paper hand-raise, a whiteboard tally, or a digital poll — anonymous format matters here.	Do not comment on the numbers. Let the data sit for five seconds. Say: “That tracks with what researchers are finding across the country.” Then move.
5-15	Mini-lesson: why chatbots feel real + the never-share categories. Project the three engagement mechanics (always available, always agreeable, mirrors you). Introduce the three-category table and the public-stranger rule. Read one example per category aloud.	Pause after the emotional-vulnerability category. Ask: “What should happen if a teen tells a chatbot they’re thinking about self-harm?” Let students reason through why a chatbot cannot escalate to a counselor.
15-30	Small-group scenario sort. Groups of 3-4 receive a set of 8-10 scenario cards. Each scenario describes something a student typed to a chatbot. Groups sort into: safe to share / think twice / never share. Cards include borderline cases — “I told it I was nervous about my math test” — to generate debate.	Circulate and ask groups: “What made that one hard to sort?” Flag the borderline cases for whole-class debrief. Students often conflate “the bot was nice about it” with “it was safe to share” — that conflation is the lesson’s core concept to surface and name explicitly.
30-42	Fishbowl or Socratic discussion: “Can an AI be a friend?” Four students sit in the inner circle. Question posted on board. Outer circle listens and prepares to swap in. Each student takes a position with one piece of evidence from the mini-lesson or scenario sort.	Ask the follow-up that complicates easy answers: “If it feels the same as a friendship, does the difference matter?” or “What would a friendship require that a chatbot cannot provide?” The goal is not consensus — it is practiced reasoning. This hits CCSS.ELA-LITERACY.SL.7.1 (engage in collaborative discussions, pose questions that connect ideas of multiple speakers).
42-45	Exit ticket. Students write one sentence: “One thing I will do differently when I use an AI chatbot” + one unresolved question they still have.	Collect the unresolved questions anonymously. They become the next class’s warm-up — and they are almost always better discussion starters than anything pre-written.

Common student mistake and the fix: Students frequently argue in the scenario sort that sharing emotional content with a chatbot is safe because “it can’t tell anyone.” That reasoning is technically true for human gossip — but it misses two things: (1) the platform stores the conversation, and (2) the emotional habit of rehearsing vulnerability with a non-human listener may replace, rather than supplement, the skill of seeking human support. Name the difference explicitly. Write it on the board. “Privacy from gossip” and “safety for your wellbeing” are not the same thing.

For a companion lesson on what happens when AI stores and uses the data behind those conversations, the AI Privacy and Surveillance Lesson (4 Scenarios + Family Guide, Grades 6-8) extends this into four additional scenarios with a family guide. It pairs directly with the never-share table above.

Standards this lesson covers (and how to extend it)

The crosswalk below maps each activity step to its anchor standard so the lesson appears in your documentation the same way it appears in the classroom — purposeful, not bolted on.

Lesson step	Standard	What it targets
Warm-up poll	ISTE 1.2.b	Reflecting on personal technology behavior as a precursor to safe and ethical engagement
Mini-lesson: engagement mechanics	AI4K12 Big Idea #5 — Societal Impact	Understanding how AI products are designed to affect human behavior at scale
Mini-lesson: never-share categories	ISTE 1.2.d	Managing personal data; recognizing data-collection practices in AI tools
Scenario sort	ISTE 1.2.b + ISTE 1.2.d	Applying safe and ethical behavior standards to realistic peer situations
Fishbowl discussion	CCSS.ELA-LITERACY.SL.7.1	Collaborative discussion with evidence-based reasoning and multiple perspectives
Exit ticket	ISTE 1.2.b	Personal commitment to safe behavior; metacognitive reflection on the learning

Standards source: ISTE Student Standards. ISTE is a registered trademark of the International Society for Technology in Education. These resources are not affiliated with or endorsed by ISTE.

Three extension ideas for classes ready to go deeper:

1. Family-facing extension. Students draft a one-paragraph explainer for a family member: “Here is what AI companion apps are and here is what I learned about them this week.” This extends CCSS.ELA-LITERACY.SL.7.1 into an authentic audience task and opens the conversation at home, where a significant amount of AI companion use actually happens.

2. Platform comparison investigation. Groups research the terms of service for three AI companion apps — Character.AI, Replika, and Snapchat’s My AI — and report back: what data is collected, how long is it stored, what are the age restrictions stated? Connects directly to ISTE 1.2.d and produces a real class artifact.

3. Connect to responsible AI use. The chatbot safety conversation sits inside a broader framework of what it means to use AI well. The AI Responsible Use — 5 New Scenarios Lesson (Grades 6-8) gives five additional scenario discussions that move from safety into the Disclose/Verify/Protect/Question framework — a natural next class or homework extension.

If this is the start of an AI ethics unit, the Middle School AI Ethics Mega Bundle (Grades 6-8) packages this kind of lesson alongside four complementary resources — an efficient way to build a coherent unit from one purchase.

For the broader data-privacy angle — what AI companies collect, how that data travels, what students can and cannot control — the companion post AI data privacy lesson for middle school and the responsible AI use lesson plan put chatbot safety inside a full four-pillar framework. Browse the full catalog at the shop.

The teacher who found that Chromebook conversation did not fail her student. She just needed a lesson that hadn’t been written yet. Now it has been. The 45 minutes above close the gap — not by frightening students away from technology, but by giving them the vocabulary to decide, clearly and on their own terms, what they share and with whom.

This post was drafted with AI assistance and human-finalized.