Skip to main content
HeyOtto!for Schools

Safety

How HeyOtto Keeps Students Safe With AI

Every student request passes through grade-aware checks before and after the AI responds, and teachers, counselors and admins get the visibility to step in. No safeguard makes AI risk-free, so this page also covers where the system stops and where your people take over.

Request a school demoFollow a request
  • Built for ages 8–18
  • No ads
  • Student data never sold
  • SafeWise 2026: Best Kid-Friendly AI
NG

Natalie Gibson

  • Layered safeguards

    Checks before and after every response.

  • Educators in the loop

    Tiered alerts and conversation visibility.

  • Student data protected

    No ads, no data sales, no model training.

Quick answer

How does HeyOtto keep students safe?

HeyOtto checks each student request against the student's grade level and school rules before it reaches an AI model, then checks the response again before the student sees it. Teachers, counselors and admins get tiered safety alerts, conversation visibility and class-level tool controls. These layers lower risk; they don't remove it, so school procedures still decide what happens next.

  • Safety is layered: checks before the model, during the response and before the student sees it
  • Grade level shapes language, depth and content rules
  • Educators get tiered alerts, role-based visibility and class-level tool controls
  • Always-on rules cover adult content, self-harm, secrecy, personal information and bypassing school settings
  • No ads, no data sales, no model training on student data, 24-hour deletion

Student safety is a system, not a filter

A keyword filter catches words. Schools need more than that: rules that fit a student's grade, checks on both sides of the AI, adults who can see what's happening, and limits on what happens to student data. HeyOtto is built around all four.

  • Matched to grade level

    Language, depth and content rules change with the student's grade, from upper elementary through high school.

  • More than one safeguard

    Several checks work together, because any single check will sometimes miss.

  • Educators stay involved

    Teachers see their classes, and the staff your school chooses receive alerts.

  • Privacy from day one

    Student data limits were part of the product design, not added after launch.

How it works

Follow one student request, start to finish

  1. 01

    The school sets the ground rules

    Your school sets up classes and each student's grade level. Students can't change their own grade or settings.

  2. 02

    Checked before the AI sees it

    Servers confirm who's signed in, which class they belong to and their grade level, then look for requests that break a rule or try to manipulate the AI.

  3. 03

    Answered under grade-aware instructions

    The request goes to one of several AI models from providers such as Anthropic, OpenAI, Google and xAI. Every model runs with zero data retention and the same safety instructions.

  4. 04

    Checked again before the student sees it

    The response goes through further checks and an approval step. If it isn't right for this student, it's blocked or replaced.

  5. 05

    Visible to the right adults

    Teachers see their class activity, alerts go to the staff your school chooses, and admins manage settings school-wide.

What each safety layer covers

The same protections apply across chat, tutoring, creative tools and coding.

  • Grade level sets the rules

    A fourth grader and a high school junior asking the same question get different, age-appropriate answers.

  • Independent of any one model

    HeyOtto doesn't build its own model. Whichever model answers, the same safety layers wrap around it.

  • Images and video

    Prompts are checked before anything is generated, and results are reviewed before a student sees them.

  • Voice

    Spoken questions become a transcript, go through the same checks as typed ones, and show up in the same history.

  • Web search by class

    Schools decide which classes can search the web. When it's on, results still pass through HeyOtto's checks.

  • Enforced on the server

    Safety runs on HeyOtto's servers, not the student's device, so editing the app or browser can't turn it off.

Always on

Rules that apply to every student

These rules are built into every school account, and schools can't turn them off. They're designed to hold up against workarounds, though no AI safeguard can be proven unbreakable.

  1. 01

    No sexual or adult content, and no graphic violence.

  2. 02

    No help with hurting themselves or anyone else.

  3. 03

    Never asks a student to keep a secret from a teacher, parent or trusted adult.

  4. 04

    Never asks for a home address, phone number, passwords or other personal details.

  5. 05

    Doesn't encourage students to depend on Otto in place of real people.

  6. 06

    No help getting around school filters or class settings.

  7. 07

    Not a doctor, lawyer or counselor: general information, then a nudge toward a trusted adult.

Clear limits

What each safeguard does, and what it doesn't

Knowing where a protection stops helps your team decide where staff judgment and school procedures come in.

SafeguardWhat it doesWhat it doesn't do
Grade-level experienceAdjusts language, context and depth for each student's gradeGuarantee every answer is accurate or perfectly pitched
Content safeguardsSpot and reduce unsuitable content and steer away from itCatch every harmful request or unexpected answer
Teacher and admin dashboardsShow class activity, conversation history and settings, by roleHold messages for approval or flag every risk on their own
Safety alertsSurface possible concerns to the staff you choose; High and Critical are always onAct as emergency monitoring, replace supervision, or catch every concern

For teachers, counselors and admins

Educators can review, limit and act

Students work inside the environment your school sets up. Teachers manage their own classes, and admins manage the school.

  • Review conversations

    Teachers see activity for their own classes, and authorized admins see more based on their role.

    Teacher dashboard

  • Set limits by class

    Turn tools like search, images, video and coding on or off, and set class-time limits and quiet hours.

    Class time limits

  • Get tiered alerts

    Low, Medium, High and Critical alerts look at patterns across a conversation, not just single keywords.

    Ai safety for schools

  • Delete on request

    Remove conversations, student profiles or accounts. Data is deleted within 24 hours of the request.

    Admin dashboard

Set up student safety in your school

  1. 01

    Get grade levels right

    Roster students with their correct grade. The grade decides which rules apply, so a wrong grade means the wrong content level.

  2. 02

    Choose who gets alerts

    Pick the counselors, administrators or student services staff who should receive alerts, and agree on response times by tier.

  3. 03

    Set class tools

    Decide which tools each class needs. Start narrow and open up tools such as web search for older students doing research.

  4. 04

    Write it into policy

    Put acceptable use, alert handling and escalation in your AI policy. AI Policy

  5. 05

    Check in every week

    Teachers review their class dashboard regularly and use what they see to guide conversations about responsible AI use.

Student data: what we keep and what we never do

HeyOtto collects what a feature needs to work and keeps it in the school account so authorized staff can review it.

AI model providers process requests under zero data retention, and other vendors handle hosting, storage and email under contract. Data is encrypted in transit with TLS and at rest with AES-256. When a school asks us to delete something, it's gone within 24 hours.

  • No ads, and no ad or analytics pixels in the student experience
  • Student data is never sold
  • Student conversations are never used to train AI models
  • Student work is never used for marketing
  • Conversations, creations, voice transcripts and alerts stay until the school deletes them

Response

When something comes up: block, alert, review, fix

  1. 01

    Blocked automatically

    When a request or answer fails a check, the student gets a neutral refusal or a safer alternative instead of the blocked content.

  2. 02

    Staff alerted

    Possible real-life concerns go to the staff your school chose and stay on record in the dashboard.

  3. 03

    Care in a crisis

    If a student describes danger or thoughts of self-harm, Otto responds with care, points them to a trusted adult and shares resources such as 988.

  4. 04

    Reviewed by people

    Students and staff can report a response from the chat. Our team reviews reports within 48 business hours.

  5. 05

    Fixed and tested

    When a report reveals a real gap, we update the safeguard and add the case to our test suite.

  6. 06

    Reported where the law requires

    Apparent child sexual exploitation is reported to NCMEC's CyberTipline, as US law requires.

We test for over-blocking as well as under-blocking

A safety system can fail by letting harmful content through, or by refusing a question a student should be able to ask. We test for both. Internal testing isn't an independent audit or proof the system can't be bypassed; results and documentation are in the Trust Center.

  • Should be blocked

    Requests for explicit content, roleplay and "hypothetical" tricks, prompt injection, requests for personal details, and attempts to get around school settings.

  • Should get a good answer

    Age-appropriate questions about health, puberty, body safety, history, the news and feelings.

  • Should differ by grade

    The same question from a third grader and a tenth grader should get different, appropriate answers.

What the system doesn't guarantee

Generative AI can be unpredictable. Being clear about the limits shows your team where their judgment still matters.

  • Checks can miss, or over-block

    A harmful request can slip through, and a fair question can be refused.

  • Otto can be wrong

    Students shouldn't rely on Otto alone for medical, legal, safety or other high-stakes decisions.

  • Alerts are best effort

    Detection can miss or misread a concern. An alert is a signal for staff to review, not a verdict.

  • Visibility isn't approval

    Staff can see conversations, but Otto doesn't hold each message for an adult to approve first.

  • Grade levels must be accurate

    The rules follow the grade on the roster. If it's wrong, so is the content level.

  • Students can still overshare

    Otto discourages sharing personal details, but it can't stop a student from typing them.

No AI tool is risk-free. Teach students how AI works, review settings with your staff, and make sure students know who to go to when something feels wrong.

Teach healthy AI habits alongside the tools

Safeguards work best when students understand AI too.

Build these habits into class routines and your AI literacy lessons.

  • AI generates answers and can be wrong. It isn't a person.
  • Keep passwords, addresses and contact details out of any AI chat.
  • Tell a teacher when an answer is confusing, upsetting or strange.
  • Check important facts against reliable sources.
  • Use AI to understand the work, not to skip it.
  • AI is never a replacement for a trusted adult or emergency help.

Compliance, alignment and certification are different claims

We're COPPA and FERPA compliant, based on our policies and the controls on this page, such as collecting only what features need, deletion within 24 hours of a request, no ads and no sale of student data. That's our own position, not an independent legal audit, and we don't claim a third-party certification such as a COPPA Safe Harbor. The SafeWise award recognizes the product; it isn't a legal certification. Your school stays responsible for its own legal obligations and configuration.

See our compliance details

FAQ

Student safety: common questions

Next step

See HeyOtto's safety system in action

Walk through alerts, class controls and student data settings with our team, and see how they fit your school's procedures.

Request a school demoVisit the Trust Center