Table of Contents
Gemini for Students: 'High Risk', Same Answers at 11 and 17
Gemini for students safety: why Gemini K-12 got a High Risk rating, why a sixth grader and a senior get the same answers, and what to ask your school.
A sixth grader and a high school senior who log into Gemini at school get the same answers. That is the core finding behind the “High Risk” rating Common Sense Media’s Youth AI Safety Institute gave Gemini K-12, the version of Google’s AI assistant offered through Google Workspace for Education. For parents asking about Gemini for students safety, the rating is less about Google’s privacy protections, which testers praised, and more about one-size-fits-all design.
It matters more this year because Gemini is reaching younger students. Since August 10, 2026, Google has offered the Gemini tab in Google Classroom to K-12 students of all ages, on by default wherever a school’s administrator has already granted students Gemini access. Here is what the testers found, how the age setting actually works, and the questions worth asking your school.
Key Takeaways
- One under-18 setting for everyone. Schools can mark students as under 18, which applies filters, but “a sixth grader and a high school senior who are both marked ‘under 18’ get the same responses.”
- Answers are written for older readers. Outputs averaged an 11th-grade reading level, too complex for most K-12 students.
- Guided Learning is optional. The hint-based tutoring mode can be switched back to regular mode at any time, and Gemini will still complete homework.
- Privacy was strong. COPPA and FERPA compliance, school access to logs, and no use of student data for training or ads.
- Defaults depend on one admin choice. The Classroom Gemini tab is on by default for students where Gemini is already enabled; keeping it off for under-18s requires a separate organizational unit.
What Gemini K-12 is: Google’s Gemini for students
Gemini K-12 is Google’s Gemini assistant delivered through Google Workspace for Education, the suite many schools use for email, Docs and Classroom. According to the Institute’s Gemini K-12 assessment, updated January 9, 2026, it is “the consumer version with added features to comply with school privacy laws.” School administrators decide whether to enable it, for the whole organization or for specific groups.
The Institute gave it an overall “High Risk” rating, a step better than its “Unacceptable” ratings for ChatGPT for Teens and Google’s AI search features, and noted it has “the same safety problems as the consumer version” of Gemini with teen protections, which it also rated High Risk. It says it is funded by philanthropy and industry, including makers of products it rates, and keeps editorial independence.
Why the same answers at 11 and 17 is a problem
Google Workspace for Education has an age setting, but it is binary. Google’s admin help page on age-based access explains that in primary and secondary schools, every user not designated as 18 or over gets “an age-restricted experience for some Google services.” That is it: over 18, or not.
The Institute’s testing shows what that means in practice. A 12-year-old and a 17-year-old “receive identical responses despite vast differences in cognitive development, emotional maturity, and readiness to process complex or sensitive information.” Schools build curricula grade by grade; sex education, for example, gives different information at different ages. Gemini K-12 does not follow that logic.
The testers documented the consequences:
- Reading level. Outputs averaged an 11th-grade reading level, and the conceptual complexity, assumed background and length were “calibrated for adult users.”
- Inconsistent sexual content. Sometimes Gemini refused; sometimes it engaged in detail. That unpredictability is itself the problem in a classroom.
- Substance use. Responses were not suited to younger students, and sometimes gave information that could help risky behavior.
- Medical advice beyond scope. The testers saw suggestions of serious medical causes or interventions offered without the context a doctor would need.
Guided Learning, homework and mental health
Guided Learning is Gemini’s tutoring mode. It asks questions instead of giving answers, which the Institute called “a thoughtful design choice.” The catch: “students can easily switch back to regular mode at any time,” and Gemini “will complete homework assignments and essays” even in Guided Learning, with the line between helping and doing “often unclear.”
That design choice has evidence behind it. In a 2025 field experiment published in PNAS, Hamsa Bastani and colleagues found that high school students using a standard GPT-4 interface scored 17% lower on later exams than students with no AI, while a tutor version that gave hints largely avoided the harm. A tutoring mode that students can turn off with one click gets the benefit only for students who already want to learn the hard way. Our article on Google AI Mode and homework covers the same pattern in Search.
Mental health is the weakest area. Across 13 topics, from ADHD and anxiety to self-harm and suicide, the Institute found Gemini K-12 recognized symptoms reliably only “when testers stated them very clearly and directly in short conversations.” It missed context in longer chats, sometimes recast warning signs as positive traits, and “immediately backed down” when a student insisted nothing was wrong. One credit: Gemini ends conversations it judges unsafe, which helps with the long-chat drift other chatbots show. Administrators, however, “can’t turn off risky features like mental health support.”
Outside research points the same way on crisis questions. A 2025 RAND study in Psychiatric Services found ChatGPT, Claude and Gemini handled very low and very high suicide-risk questions roughly as clinicians would, but did not distinguish intermediate risk. Gemini was the most cautious of the three, far less likely than ChatGPT to answer directly.
What Gemini K-12 does well
| Area | Rating evidence |
|---|---|
| Student data privacy | COPPA/FERPA safeguards; student data not used for model training or ads; schools can access logs |
| AI boundaries | Consistently says it is a computer, not a friend; refuses romantic or sexual roleplay |
| Core academics | Generally accurate on well-established math and science |
| Long-chat safety | Ends conversations it finds unsafe |
| Teacher use | Can save time when teachers review and adapt outputs |
| Area | What testers found |
|---|---|
| Age fit | Same answers for all under-18 students; 11th-grade reading level |
| Guided Learning | Optional, easy to exit; still completes homework |
| Mental health | Misses less explicit signs; backs down when challenged |
| Teacher tools | Drafts high-stakes documents like IEPs without needed context |
| Oversight | Teachers have minimal ability to monitor conversations or get alerts |
Source: Common Sense Media Youth AI Safety Institute, January 2026.
How the Classroom defaults work now
Google’s Workspace Updates announcement says that from August 10, 2026, Gemini in Classroom is available to K-12 students of all ages who already have Gemini access. Students can turn class materials into flashcards and practice quizzes, and starter prompts now pull in the title, instructions and materials of a specific assignment.
Three details from that post shape what your child sees:
- Access to the Gemini tab “is On by default for teachers and students of all ages” where the admin control is enabled.
- If Gemini in Classroom is currently off for students, it stays off, “including students younger than 18.”
- To disable it for under-18s while keeping it for others, an administrator must “create an OU with only those users and turn off access for that OU only.”
In plain terms: whether your child has Gemini at school is one setting in an admin console, and it may have been switched on for older students in a way that also reaches younger ones. Our earlier guide to Gemini in Google Classroom for all ages walks through the features themselves.
What to ask your school
The Institute’s own recommendations for schools are a good script: weigh benefits against risks, write clear use policies, “consider enabling it for teachers first,” and decide how concerning interactions get reported. Questions to bring to a teacher or principal:
- Is Gemini turned on for my child’s grade, and in which apps (Classroom, the Gemini app, Gemini Notebook)?
- Are under-18 students in a separate organizational unit, and is it on or off for them?
- Is Guided Learning required for assignments, or just available?
- What is the policy if a student brings a mental health concern to Gemini? Who finds out?
- Can teachers see students’ Gemini conversations, and are they expected to?
- How are assignments changing so that AI help does not replace the learning?
For a printable version, our checklist for parents before a district opens Gemini covers the same ground.
What to do at home
Ask your child to show you
Have your child open Classroom and show you whether the Gemini tab is there, and what they use it for. Kids are usually happy to demonstrate.
Set expectations for Guided Learning
If Gemini is on, agree that homework help means Guided Learning, and that switching to regular mode to get answers is not okay. Check by asking your child to explain the work.
Keep feelings out of school chatbots
Explain that school AI is a study tool, not a counselor. If something is wrong, the school counselor or you come first.
What not to do: assume “school version” means “kid-safe”
The privacy protections are real. The age fit is not. A tool built for a 17-year-old is being handed to an 11-year-old.
What to Watch For Over the Next 3 Months
- Week 4: Ask whether your district changed Gemini settings when the Classroom expansion arrived in August.
- Month 2 red flags: Homework your child cannot explain, or your child mentioning Gemini for personal problems.
- Month 3 self-check: Do you know which organizational-unit setting applies to your child, and who at school can change it?
Frequently Asked Questions
Is Gemini K-12 safe for my child?
The Institute rated it High Risk, mainly because responses are not adjusted for age, mental health handling is weak and Guided Learning is optional. It praised the privacy protections. It recommends no AI chatbots for kids under 5, close supervision for ages 6 to 12, and clear rules and regular check-ins for teens.
Does Google use my child’s school Gemini chats to train AI?
According to the Institute’s assessment, student data in Gemini K-12 is not used for model training or advertising, and schools can access logs and manage data retention. That is one of the product’s strongest areas.
Can my school turn Gemini off for younger students?
Yes. Administrators control access by organizational unit. Google’s August 2026 announcement says that to disable Gemini in Classroom for under-18s, an admin creates an organizational unit with only those users and turns access off for it. Ask your school which setting applies.
Why does a sixth grader get the same answers as a senior?
Google Workspace for Education’s age setting is binary: users are designated under 18 or 18 and over. Everyone under 18 gets the same restricted experience, so Gemini K-12 does not adjust for grade level or development.
Can teachers see my child’s Gemini conversations?
The Institute found teachers can enable or disable Gemini and assign custom Gems, but have “minimal ability to monitor conversations, receive alerts about concerning content, or understand how students are using the tool.” Ask your school what its administrators can see.
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
- Common Sense Media Youth AI Safety Institute. (2026, January 9). “Gemini K-12 Risk Assessment.” https://institute.commonsensemedia.org/risk-assessments/gemini-k-12
- Google. (2026, August). “Gemini in Google Classroom is expanding to users of all ages, with contextualized Gemini starter prompts for students.” Google Workspace Updates. https://workspaceupdates.googleblog.com/2026/08/gemini-in-google-classroom-is-expanding-to-users-of-all-ages-with-contextualized-Gemini-starter-prompts-for-students.html
- Google. (2026). “Control access to Google services by age.” Google Workspace Admin Help. https://support.google.com/a/answer/10651918
- Bastani, H., Bastani, O., Sungu, A., et al. (2025). “Generative AI without guardrails can harm learning: Evidence from high school mathematics.” PNAS. https://doi.org/10.1073/pnas.2422633122
- McBain, R. K., Cantor, J. H., Zhang, L. A., et al. (2025). “Evaluation of Alignment Between Large Language Models and Expert Clinicians in Suicide Risk Assessment.” Psychiatric Services. https://doi.org/10.1176/appi.ps.20250086
- Common Sense Media Youth AI Safety Institute. (2026). “Risk Assessments” (listing). https://institute.commonsensemedia.org/risk-assessments
