Table of Contents
Trust and Safety Career: The People Behind Teen AI Modes
ChatGPT for Teens shipped August 2026 with defaults someone had to design. The trust and safety career is that job. Here is the path and honest salary data.
A parent can link their teenager’s ChatGPT account, set quiet hours, and receive safety notifications. They cannot read the conversations. Somebody decided that, and had to be able to defend it to a regulator, a journalist, and an angry parent, on the same day. ChatGPT for Teens launched on August 18, 2026 with age-appropriate protections on by default and extra safeguards in self-harm, violence, eating disorders, dangerous activities, and explicit content. Every one of those is a line someone drew. The trust and safety career is the job of drawing them, measuring whether they worked, and moving them when they didn’t.
Key Takeaways
- ChatGPT for Teens rolled out globally to eligible accounts on both free and paid personal plans starting August 18, 2026, with age-appropriate protections on by default and extra safeguards in self-harm, violence, eating disorders, dangerous activities, and explicit content.
- The parental controls are deliberately partial: guardians can link an account, set quiet hours, enable Study Mode by default, and receive safety notifications, but parental controls do not let a parent read or monitor the teen’s conversations.
- Product decisions carry the safety design: when a teen appears to be attempting to cheat, the system nudges toward Study Mode rather than refusing outright. OpenAI says the protections are informed by developmental science and expert guidance.
- The regulatory backdrop is real. The FTC issued 6(b) orders to seven companies in September 2025 over chatbots and minors, and California’s SB 1119, signed September 10, 2026, mandates age determination, default session limits, disabled memory, and independent child-safety audits every two years for companion chatbots.
- Honest pay picture: no BLS code exists for trust and safety. The closest categories (May 2025) are management analysts at $101,860 with 10% growth, information security analysts at $129,180 with 21%, psychologists at $99,110 with 6%, and data scientists at $120,230 with 35%.
What trust and safety actually means
Trust and safety is the function inside a technology company that decides what the product will not do, builds the systems that enforce it, and measures whether the enforcement works. It sits between policy, engineering, data science, legal, and operations, and it reports to whoever the company thinks is accountable when something goes wrong.
For a teen AI product, the decisions look like this.
Who counts as a teen. OpenAI places users aged 13–17 in the teen experience, using self-reported age and age prediction signals. Every age system has two failure modes: an adult misclassified as a teen loses functionality, and a teen misclassified as an adult loses protection. Someone chooses which error to bias toward.
What the defaults are. Protections on by default, in five named harm categories. Defaults matter more than options because most people never change them.
What parents can and cannot see. This is the hardest call in the whole product. Full transcript access would reassure parents and would also push teens to a second account. No access preserves the relationship and leaves parents with less information. OpenAI chose the second, with safety notifications as the compromise.
What happens at the edge. A teen mentions self-harm. A teen asks a question that is fine for an adult and not for a 14-year-old. A teen tries to submit an AI-written essay. Each case has a designed response, and the cheating case is instructive: a nudge toward Study Mode rather than a refusal, which is a bet that redirection works better than blocking.
The evidence supports that bet, weakly. Common Sense Media found in August 2026 that 44% of teens had a tool blocked at school and 59% of those switched to a personal device. Blocking relocates behaviour. Whether nudging changes it is a question the field is still answering.
Why the regulators are the other half of this job
Two developments make trust and safety a compliance function as well as a design one.
The FTC’s 6(b) inquiry, issued September 11, 2025 to Alphabet, Character Technologies, Instagram, Meta, OpenAI, Snap, and X.AI, demanded information on how companies monetize engagement, develop characters, test for harm to minors, and handle personal information from conversations. That is a list of things someone inside each company now has to be able to document.
California’s SB 1119, approved September 10, 2026, goes further for companion chatbots specifically: age determination before a child gets access, defaults that disable persistent memory and push notifications and cap sessions at one hour continuous and two hours daily, disclosure reinforced during long interactions, documented risk assessments before launch, independent child-safety audits every two years, and crisis protocols with parent notification.
Read that as a job description. Risk assessments, audits, age assurance, documented crisis protocols. Those are trust and safety deliverables, and they are now legally required rather than optional. Our companion piece on what SB 1119 changes for families covers the parent side.
Role table: the jobs inside a trust and safety org
| Role | What they do day to day | Typical background | Where they sit |
|---|---|---|---|
| Policy manager | Writes the rules: what the product allows, for whom, with what exceptions | Law, public policy, or journalism | Policy team, reports to legal or product |
| Child safety specialist | Owns the under-18 experience: defaults, age assurance, crisis flows | Child development, social work, psychology | Trust and safety, often with clinical advisors |
| Trust and safety engineer | Builds classifiers and enforcement systems | Software engineering, ML | Engineering, embedded with policy |
| Data scientist, integrity | Measures whether a policy worked: false positive rates, harm incidence, appeal outcomes | Statistics, data science | Analytics |
| Investigations analyst | Handles escalations, coordinated abuse, law enforcement requests | Criminal justice, intelligence, journalism | Operations |
| Content moderator | Reviews flagged material at volume. High turnover and real psychological cost | Varies; entry level | Operations, often outsourced |
| Clinical advisor | Reviews crisis flows and self-harm handling against clinical standards | Clinical psychology, psychiatry, counseling | Advisory, sometimes contracted |
| Regulatory compliance lead | Produces the risk assessments and audit evidence that laws like SB 1119 require | Law, compliance, audit | Legal |
| Researcher, youth | Studies how teens actually use the product, including workarounds | Developmental psychology, human-computer interaction | Research |
Two things to say plainly to a teenager considering this. The content moderator row involves reviewing disturbing material and has documented psychological costs and high turnover; it is a real entry point into the field and it should be entered with open eyes. And the clinical advisor row is why a psychology degree, median $99,110, is a legitimate route into a technology company.
The trust and safety career path: school to a working desk
| Stage | What to take or earn | Cost and time | Entry roles | What that role does daily |
|---|---|---|---|---|
| High school | Psychology if offered, statistics, government, English composition; debate; moderating a school club’s online space | Free | Peer moderator, student council | Handling conflict between people who both think they’re right |
| Bachelor’s | Psychology, sociology, public policy, criminal justice, communications, or computer science. Statistics is the common requirement across all of them | 4 years | Content moderator, operations analyst, policy assistant | Reviewing cases, applying rules, escalating edge cases |
| Certification | No licence. Useful: privacy credentials such as CIPP, child-safeguarding training, data analysis certificates | Weeks to months | Policy analyst, compliance analyst | Documenting decisions, producing audit evidence |
| Master’s | Public policy, social work (MSW for the clinical track, required for clinical licensure), human-computer interaction, or data science | 1–2 years | Child safety specialist, policy manager, integrity data scientist | Owning a policy area end to end |
| Doctorate (advisory track) | Clinical or developmental psychology. Licensure requires supervised experience and state licensing in most U.S. states | 5–7 years | Clinical advisor, research lead | Reviewing crisis flows, publishing, advising product |
| Technical route | Computer science, then ML | 4 years | Trust and safety engineer | Building classifiers, enforcement tooling |
| Day to day | Write policy, argue about edge cases, measure outcomes, respond to regulators, explain decisions to people who hate them | — | — | — |
Pay and demand, said plainly
| Closest official category | Median pay (May 2025) | Projected 2025–2035 | Jobs (2025) |
|---|---|---|---|
| Information security analysts | $129,180 | +21% | ~192,900 |
| Data scientists | $120,230 | +35% | ~275,600 |
| Management analysts | $101,860 | +10% | ~1,080,000 |
| Psychologists | $99,110 | +6% | ~209,400 |
| Social workers | $61,780 | +6% | ~816,100 |
There is no BLS occupation code for trust and safety, so anyone giving you a median for it is estimating from job postings. What can be said honestly: policy and program roles at large technology companies are typically compensated near or above the management analyst median, technical integrity roles near the data scientist and security analyst medians, and content moderation, often outsourced, well below all of them.
The demand direction is clear even without a precise number. Every new regulation, the FTC inquiry, SB 1119, the state companion-chatbot wave across roughly a dozen states, the EU transparency obligations, creates documentation, audit, and assessment work that did not exist before. That is structural growth. The honest caveat is that trust and safety teams have also been cut during technology industry contractions, so it is not a recession-proof function.
What a 10–15-year-old can do this year
Read a product’s actual safety settings
Open ChatGPT for Teens, Discord’s safety settings, or Roblox’s parental controls with your kid and read what each toggle actually does. Then ask which default they’d change and why. That is the core exercise of the job.
Find the workaround, then don’t use it
Ask your teenager how a determined 14-year-old would get around a given protection. They will know. Then talk about what a designer could do instead. Thinking like an adversary without behaving like one is the defining professional skill here.
Write the parent-access policy
Give them the real question: should parents be able to read a teenager’s AI chats? Have them write a one-page policy with reasons, then argue the opposite side. OpenAI chose no access plus safety notifications. Whether that’s right is a genuine open question, and a 15-year-old has standing to have an opinion.
Measure something
Take any rule at their school and estimate how many students it affects, how many comply, and how you’d find out. Trust and safety runs on measurement, and most people skip straight to opinion. Our guide to setting up ChatGPT for Teens in 30 minutes gives you a concrete artifact to examine together.
What not to do
Don’t romanticise it. This field involves reading things you cannot unread, being blamed by everyone, and shipping compromises. Content moderation in particular carries documented psychological cost. A kid who is drawn to it because they want to protect people should know that the job is mostly writing documentation and arguing about definitions, and that the protection happens through those boring artifacts.
What to Watch For Over the Next 3 Months
- Week 4: Your kid can name one default setting in a product they use and explain what it protects against.
- Month 2 red flags: They only want to find loopholes. They think every safety feature is either perfect or pointless. They can’t articulate a reason a company might restrict something beyond “they don’t want to get sued.”
- Month 3 self-check: Ask them to defend the decision not to give parents transcript access, then to attack it. If they can do both credibly, they have the disposition this work requires.
Frequently Asked Questions
Is trust and safety a growing field?
Structurally yes, driven by regulation: the FTC 6(b) inquiry, California’s SB 1119 with its mandatory risk assessments and biennial audits, and the state companion-chatbot wave all create work that legally must be done. The caveat is that these teams have also been cut during industry downturns, so it is growing and not recession-proof.
Does my kid need a psychology degree?
No, but it’s one of several good routes, particularly for child safety and clinical advisory roles. Psychologists earn a median of $99,110 and most states require licensure for clinical practice. Policy, data science, and engineering routes are equally common inside these teams.
Is content moderation a reasonable first job?
It is a real entry point and it carries documented psychological cost, with high turnover in the industry. If your kid pursues it, they should ask about wellness support, rotation policies, and whether the role is in-house or outsourced before accepting.
Why don’t parents get to read the conversations?
Because the designers judged that full surveillance would push teenagers to unmonitored accounts, losing the protections entirely. Common Sense Media found 59% of teens blocked at school switched to personal devices, which is the evidence behind that reasoning. It’s a defensible call, not an obviously correct one.
What if my kid is technical and wants to build rather than write policy?
Trust and safety engineering is a full career: building classifiers, enforcement pipelines, and the measurement systems that tell you whether a policy worked. It pays near software and security medians and sits inside engineering while working daily with policy people.
About the author
Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.
Sources
- OpenAI. (2026). “ChatGPT for Teens.” Help Center, August 18, 2026. https://help.openai.com/en/articles/20001421-chatgpt-for-teens
- OpenAI. (2026). “Our approach to age prediction.” https://openai.com/index/our-approach-to-age-prediction/
- TechCrunch. (2026). “OpenAI launches a safer ChatGPT for teens, years after teens started using it.” August 18, 2026. https://techcrunch.com/2026/08/18/openai-launches-a-safer-chatgpt-for-teens-years-after-teens-started-using-it/
- California Legislature. (2026). “SB-1119 Companion chatbots: children’s safety.” Approved September 10, 2026. https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=202520260SB1119
- Federal Trade Commission. (2025). “FTC Launches Inquiry into AI Chatbots Acting as Companions.” September 11, 2025. https://www.ftc.gov/news-events/news/press-releases/2025/09/ftc-launches-inquiry-ai-chatbots-acting-companions
- Common Sense Media. (2026). “Teens in the AI Era: Schoolwork and the Skills That Matter.” August 18, 2026. https://www.commonsensemedia.org/research/teens-in-the-ai-era-schoolwork-and-skills-that-matter
- Bureau of Labor Statistics. (2026). “Psychologists.” Occupational Outlook Handbook, May 2025 data. https://www.bls.gov/ooh/life-physical-and-social-science/psychologists.htm
- Bureau of Labor Statistics. (2026). “Management Analysts” and “Social Workers.” May 2025 data. https://www.bls.gov/ooh/business-and-financial/management-analysts.htm
