AI Safety Audit: What ChatGPT for Teens Means for Your Business
OpenAI launched ChatGPT for Teens on August 18, 2026, a dedicated experience for users aged 13 to 17 with safety protections switched on by default [1][3]. Anyone the system estimates is under 18, or who states they are 13 to 17, is automatically placed into the teen experience [1]. Age estimation, content guardrails, parental controls for AI, and privacy limits are no longer optional extras at the largest AI company in the world. They are the shipping default.
That is a signal for your business even if you don't sell to teenagers. If you deploy a chatbot, an AI assistant, or an automated support agent that the public can reach, "what are you doing for under-18 users?" is now the baseline question. OpenAI tracks more than 2,000 signals to estimate a user's age and turn on teen protections automatically [5]. It restricts high-risk conversations about self-harm, violence, eating disorders, dangerous activities, and explicit sexual or graphic content, and its model rules forbid romantic language, encouraging emotional dependence, or implying the chatbot has feelings or consciousness [1][6]. Parents with linked teen accounts can set Quiet Hours, manage selected settings, and receive safety notifications in limited high-risk situations [1].
Can your AI deployment document any of that? If a minor used your tool today, could you name your age-assurance method, your content guardrails, your data retention rules for under-18 users, or your escalation path for self-harm signals? Most small and mid-size businesses cannot. That is the gap an AI safety audit closes.
Do not assume a vendor's announcement covers your obligations either. Child-safety group Fairplay warned that OpenAI's launch could give parents a false sense of safety, noting that restricting ChatGPT's long-term memory is not enabled by default in teen accounts and would require parental-control intervention [6]. If OpenAI itself has defaults that child-safety advocates question, your deployment almost certainly has gaps worth finding before a regulator, a parent, or a news cycle finds them first.
The fix is not panic. It is an audit. Use the checklist below, then run the free readiness check in our AI audit tool, or book a deeper AI safety compliance audit if you ship consumer-facing AI and want the paperwork to prove you are ready.
Checklist: is your AI safe for minors?
1. Age verification
- You collect age at signup and enforce a minimum age (13+ under COPPA-style rules).
- You have an age-assurance mechanism for users who don't state their age honestly (behavioral signals, not just a checkbox).
- Users detected or identified as under 18 get a distinct experience, not the adult default.
- You capture verifiable parental consent before a 13-17 user gets full access.
- You document your age-assurance method, the signals you rely on, and its known limits, in writing.
2. Parental controls
- Parents or guardians can link their teen's account.
- You offer Quiet Hours or access windows parents can set.
- You send safety notifications for limited high-risk situations (e.g., possible self-harm).
- The opt-in model is documented and works from both sides (parent and teen).
- Parents can restrict features that carry risk, such as long-term memory.
3. Data privacy for under-18 users
- You know exactly what data you collect from minors and why.
- Long-term memory or profile persistence is off by default for under-18 accounts.
- You warn before sensitive or private image uploads.
- Your privacy policy has a minors section covering collection, retention, and deletion.
- You have a deletion path for a parent or guardian acting on a minor's behalf.
4. Safety and moderation layers
- Content guardrails for self-harm, violence, eating disorders, and sexual or graphic content are on by default.
- The model is prohibited from romantic language, emotional-dependence cues, or claiming feelings or consciousness.
- Break reminders and healthy-use cues exist for younger users.
- You have a documented escalation path for high-risk signals, including human review.
- Your under-18 protections are tested with teen-specific evaluations, and the results are written down.
Frequently asked questions
Does ChatGPT for Teens affect my business if I don't sell to teenagers?
Yes. OpenAI made under-18 protections the shipping default for the largest consumer AI platform. If you deploy a chatbot, AI assistant, or automated support agent the public can reach, "what are you doing for under-18 users?" is now the baseline question for any AI safety audit, whether or not you target minors.
Are ChatGPT for Teens' parental controls on by default?
No. Parental controls require a two-sided opt-in from both the teen and a parent or guardian. The safety protections in higher-risk content areas are on by default, but child-safety group Fairplay notes that restricting long-term memory is not enabled by default and would require parental-control intervention.
What should an AI safety audit check for under-18 users?
Four areas: age verification (do you collect age, enforce a minimum, and route under-18 users to a distinct experience), parental controls (account linking, Quiet Hours, notifications, two-sided opt-in), data privacy for minors (what you collect, retention, deletion paths), and safety/moderation layers (content guardrails, model-behavior rules, escalation paths, teen-specific testing).
Not sure where your business stands? Run the free AI audit tool — a ten-minute check of your permissions, data access, and the gaps worth fixing before you buy any AI service.
Sources
- OpenAI — Introducing ChatGPT for Teens (Aug 18, 2026): openai.com/index/chatgpt-for-teens
- CNBC — OpenAI rolls out ChatGPT for Teens (Aug 18, 2026): cnbc.com (Aug 18, 2026)
- The New York Times (via Boston Globe wire) — OpenAI introduces ChatGPT for Teens (Aug 18, 2026): bostonglobe.com (Aug 18, 2026)
- Associated Press (via The Star) — OpenAI introduces ChatGPT for Teens, promising a more age-appropriate chatbot (Aug 18/19, 2026): thestar.com.my (Aug 18/19, 2026)
Accuracy note: Product details (age 13-17, auto-enrollment via age estimation or stated age, 2,000+ signals, on-by-default content restrictions, model rules against romantic language, Quiet Hours, two-sided parental opt-in, safety notifications) are as announced by OpenAI on Aug 18, 2026 and corroborated by CNBC, NYT wire, and AP wire coverage the same day. The "false sense of safety" caution is from Fairplay's Brendan Bouffard, associate general counsel (AP wire); the long-term-memory-default claim is his characterization of OpenAI's teen-account settings. OpenAI does not verify ages; age estimation is signal-based and described by OpenAI as "minimally privacy invasive."