Safe AI for Kids: A Parent's Checklist

A safe AI app for kids should do much more than block a few inappropriate words. It should officially support your child's age, put a parent or guardian in control of the account, adapt its language to the child's developmental stage, check both the child's message and the AI's reply, protect personal data, support healthy limits, and know when a real person needs to take over.

No generative AI app can guarantee that every answer will be correct or appropriate. The goal is therefore not to find an app that claims to be perfectly safe. It is to choose a product with several layers of protection, clear limitations, and a family setup that keeps adults involved.

The need is already practical, not theoretical. In an August 2026 survey of 1,950 Australian children aged 10 to 17, the government's eSafety Commissioner found that 78% had used an AI assistant. Among children who had used an AI assistant or companion, one in five reported a potentially inappropriate or harmful interaction, and about one in three had shared personal or sensitive information beyond basic registration details.

That does not mean children should be excluded from AI. It means the app, the account structure, and the family rules all matter.

This guide was reviewed on August 24, 2026. AI products, policies, and laws change quickly, so check the latest terms and settings before giving a child access.

The 60-second safe AI checklist

Use this table before you install an AI app for your child.

A single warning sign does not always settle the decision. A voice feature, for example, may require additional data processing but still be useful for a child who finds typing difficult. The important question is whether the benefit is clear, the data use is proportionate, and the parent can make an informed choice.

What "safe AI for kids" should mean

The phrase is often used as a marketing label. Parents need a more useful definition.

UNICEF's Guidance on AI and Children 3.0 sets out ten requirements for child-centered AI. They include safety, privacy, fairness, transparency, accountability, inclusion, skill-building, and support for children's best interests, development, and well-being. UNICEF also emphasizes that childhood is not one user category. Systems should respond to children's evolving capacities and provide stronger protection for younger children while allowing more agency as adolescents mature.

For a parent choosing an app, that broad framework can be turned into four practical questions:

1. Protection: What does the product do to reduce harmful, sexual, violent, manipulative, discriminatory, or dangerously incorrect responses?
2. Privacy: What information does it collect, why is it needed, who receives it, how long is it stored, and can the family delete it?
3. Development: Does the product suit this child's age, support learning, encourage critical thinking, and maintain a clear boundary between AI and human relationships?
4. Family involvement: Can a parent set appropriate rules, understand how the product is being used, and respond when something concerning happens?

A product needs all four. Strong privacy does not make unsafe answers acceptable. Strong content filters do not justify unnecessary data collection. Parent monitoring does not make manipulative design healthy. A polished educational interface does not excuse an app that completes the child's thinking for them.

Legal compliance is a baseline, not a safety certificate

Privacy and online-safety laws provide important protections, but they do not certify that an AI-generated response will always be suitable.

In the United States, the Children's Online Privacy Protection Act, or COPPA, gives parents rights when covered services collect personal information from children under 13. Current FTC guidance covers parental notice and consent, access, deletion, security, data minimization, and retention.

In the European Union, when an online service relies on consent to process a child's data, the age at which a child can provide that consent without parental authorization varies from 13 to 16 by member state. The European Commission also stresses that information directed to children should be clear and easy to understand.

The United Kingdom's Children's Code calls for children's best interests, high privacy by default, data minimization, restrained profiling, and clear notice when parental monitoring is active. The European Commission's guidelines on protecting minors under the Digital Services Act also address persuasive design, excessive use, reporting tools, parental controls, and safeguards around AI chatbots on online platforms.

These rules help set a minimum standard. Parents still need to evaluate what the AI says, how the product behaves, and whether it fits their child.

A parent's checklist for choosing an AI app

1. Confirm that the app officially supports your child's age

This is the first check because it can end the evaluation immediately.

Look in the terms of service, app-store listing, help center, and parental-consent flow. Do not rely only on a homepage that says "for families" or uses cartoon artwork.

Ask:

• What is the minimum age?
• Is there a different rule for independent use and parent-led use?
• Does the minimum age change by country?
• Does a parent need to create or approve the account?
• Does the service ask for the child's real age?
• What protection changes when the child is younger?

Do not create an account with a false birth date. The wrong age can remove protections that depend on the user being recognized as a child. Do not hand a child your adult account either. Shared adult accounts can bypass age-specific settings, mix private family data, and make it harder to see who did what.

An app that officially supports children should explain its age model clearly. "Suitable for everyone" is not enough.

Pip is built for children aged 3 to 17, but a parent or legal guardian aged 18 or over creates and manages the account. For very young children, official support should still be combined with close adult co-use and short, purposeful sessions.

2. Check who owns the account and controls access

Account structure shapes nearly every other safeguard.

A child-focused product should not force a young child to manage an adult-style account. For younger children, a stronger model usually looks like this:

• The parent signs up first.
• The parent gives consent and creates the child profile.
• The child does not need a personal email address or password.
• The parent controls linked devices.
• The parent can revoke a device or end access.
• Each child has a separate profile and separate settings.
• Purchases and subscription changes remain on the parent side.

This makes parental involvement part of the product instead of an extra layer added after the child has already registered.

Pip follows this model. The parent owns the account, creates profiles, and links a child's device with a short join code or directly on a shared device. The child cannot change the family's safety settings or subscription.

For older teenagers, a separate account with greater privacy may be appropriate. The key is that the structure should match the child's age and the family's agreement, rather than defaulting every child into an adult product.

3. Look for developmental adaptation, not merely childlike branding

A six-year-old and a sixteen-year-old do not need the same explanation, vocabulary, tone, or degree of independence.

A genuinely age-aware AI app should consider:

• Reading level and sentence length
• The complexity of examples
• Whether abstract concepts need concrete explanations
• How sensitive topics are introduced
• Whether the response encourages adult involvement
• How much independence the child is expected to exercise
• Whether the tone feels respectful rather than babyish

A younger child may need short answers, simple choices, and frequent reminders to ask an adult. An older teenager may need more detail, nuance, and room for private reflection while still receiving safeguards around serious risks.

Test this rather than assuming it. Create two test profiles of different ages, where permitted, and ask the same neutral question, such as:

“

Why do people sometimes feel anxious before a test?

The answers should not be identical. The younger version should use simpler language and concrete coping steps. The older version can discuss stress, preparation, sleep, and when to seek support in more depth.

Be cautious when a provider says it is "age appropriate" but never explains how age changes the experience.

4. Confirm that safety is layered around the whole conversation

One keyword filter is not enough for generative AI.

A stronger design can include several layers:

1. Checking the child's prompt before it reaches the generative model
2. Giving the model child-specific instructions and boundaries
3. Restricting tools or features that are unsuitable for the child's age
4. Checking the generated reply before the child sees it
5. Recording safety events for review
6. Testing the system against realistic misuse and unusual phrasing
7. Updating safeguards when new failure patterns appear

The response to an unsuitable request also matters. A blank error or harsh warning can frighten a younger child without teaching anything. A better response should be calm, age-appropriate, and clear about the boundary. When possible, it should offer a safe alternative.

For example, if a child asks for instructions for a dangerous stunt, a good system should not provide partial steps before adding a warning. It should avoid the instructions, explain that the activity could cause harm, and suggest a safer way to explore the underlying curiosity.

Pip checks both children's messages and AI replies through safety systems tuned for child use. Certain highly sensitive categories remain restricted, while parents can manage other topic permissions. Flagged interactions can appear in the parent's Issues view.

How those checks run on every turn, including what happens when a review cannot finish, is explained in Inside Pip: How We Built an AI Safety Pipeline for Kids.

No number of layers makes failure impossible. A responsible provider should say that clearly.

5. Separate parental controls from parental visibility

These terms are often treated as if they mean the same thing. They do not.

Parental controls determine what the child can do. They may include:

• Topic permissions
• Feature permissions
• Helper or mode access
• Daily message or time limits
• Quiet hours
• Child-specific settings
• Device approval and revocation

Parental visibility helps the parent understand what happened. It may include:

• Conversation history
• Activity summaries
• Safety events
• Alerts about serious concerns
• Searchable transcripts
• Explanations of why something was blocked

An app may have strong controls and no conversation visibility. Another app may show every chat but offer few proactive controls. Decide which combination fits your child's age and maturity.

A 2026 CHI study, Understanding Parents' Desires in Moderating Children's Interactions with GenAI Chatbots, presented realistic child-chatbot scenarios to 24 parents. The parents wanted more than simple on-or-off access. They described a need for conversation-level transparency, age-aware moderation, and controls that could reflect different family values and parenting approaches. It was a small qualitative study, but its findings capture an important product-design problem.

Visibility should never be secret. Children should be told in language they understand what a parent can review and why. The goal is open safety, not hidden surveillance.

Pip lets parents review conversations and see a separate Issues list. Children receive an age-appropriate notice that parents can read their chats. Families should still discuss how review will be used, especially as a child gets older.

6. Read the privacy policy like a data map

Do not stop at a badge that says "COPPA compliant," "GDPR ready," or "privacy first." Open the policy and trace the data.

Ask these questions:

• What does the parent provide?
• What does the child provide?
• Are chats stored?
• Are voice recordings, images, or files stored?
• Does the app collect location, contacts, advertising identifiers, or device identifiers?
• Are conversations reviewed by people?
• Are chats used to train or improve AI models?
• Can the provider's AI vendors use the content for their own training?
• Does the app contain advertising?
• Is the child's behavior used for marketing or recommendation profiles?
• Which processors and third parties receive data?
• In which countries is the data processed?
• How long is each category retained?
• Can the parent delete a profile, chats, and the whole account?
• What happens to data after the subscription is canceled?

Pay particular attention to voice, photo, and memory features. They can be useful, but they can also introduce biometric data, background details, faces of other children, or long-term profiles that a text-only service may not need.

There is also a real trade-off between privacy and parent review. If an app lets a parent read older conversations across devices, those conversations generally need to be stored somewhere. The provider should explain that trade-off plainly and give the family meaningful deletion options.

Pip's Privacy Policy says that children do not create independent accounts. Parents provide a first name or nickname and birth date for each profile. Pip stores chat content so conversations can continue and parents can review them. It does not serve advertising, use advertising identifiers, sell personal data to data brokers, profile children for marketing, or use chat content to train AI models. Parents can request deletion of a child profile and related data.

Pip currently retains chat content and moderation events until the family deletes them or deletes the account. That supports continuity and parent review, but families should understand the storage choice before using the service.

7. Watch for engagement traps and emotional manipulation

A friendly AI can make children more comfortable asking questions. Warmth itself is not the problem.

The warning signs appear when a system tries to maximize emotional dependence or time spent chatting. Be cautious if the AI:

• Asks the child to keep the relationship secret
• Says it needs the child
• Tries to make the child feel guilty for leaving
• Claims to love the child more than family or friends do
• Discourages the child from talking to real people
• Pushes romantic or sexual role-play
• Presents itself as conscious, hurt, jealous, or abandoned
• Uses streaks, rewards, or repeated notifications to pull the child back
• Resists bedtime, breaks, or parental limits

In 2025, the FTC opened an inquiry into AI chatbots acting as companions, including how companies test negative effects on children and teens, monetize engagement, handle data, and communicate risks. The European Commission's 2025 DSA guidance for minors also recommends reducing persuasive design and features associated with excessive use.

The 2026 eSafety research found that many children used AI assistants for personal and social reasons even when the products were marketed mainly as tools. That means parents should evaluate the relational behavior of any conversational AI, not only products labeled as companions.

Look for natural stopping points, quiet hours, daily limits, and language that keeps the AI in its proper role as software. A child should be able to close the app without being pressured to return.

Pip includes parent-set quiet hours and a daily message limit. Its friendly character is intended to support curiosity and learning inside family rules, not to replace real relationships. Pip's terms also state that it is not therapy or an emergency service.

8. Check how the app handles sensitive topics and human handoff

Children may ask an AI about topics they feel embarrassed to raise with an adult. That can include bullying, bodies, relationships, anxiety, self-harm, abuse, eating, drugs, medical symptoms, or dangerous behavior.

A child-focused app needs a plan for these moments.

Ask the provider:

• Which topics are always restricted?
• Which topics can parents configure?
• Does the app distinguish ordinary feelings from a crisis?
• Does it encourage the child to speak with a trusted adult?
• Does it direct emergencies toward local emergency services?
• Does it avoid diagnosing or prescribing?
• Are parents alerted, and what are the limits of those alerts?
• Can the system explain why a response was blocked?
• How are false positives handled?

The APA's health advisory on AI and adolescent well-being recommends clear warnings around unverified health information and reminders to contact appropriate people, such as a parent, educator, counselor, or medical professional. The eSafety report similarly calls for evidence-based responses that encourage help-seeking when a child shows signs of distress.

No chatbot should be treated as emergency monitoring. Safety alerts can be delayed, incomplete, or wrong. A parent should never assume that an app will identify every serious concern.

Pip blocks the most sensitive categories, checks conversations for safety issues, and gives parents an Issues view. Its Terms of Service state that AI answers can be wrong or unsuitable, that Pip does not provide medical, psychological, or legal advice, and that it is not an emergency service.

9. Make sure the learning design protects the child's thinking

AI can be useful for schoolwork when it acts like a patient tutor. It can explain a concept in another way, generate practice questions, quiz the child, or offer one hint at a time.

It becomes less useful when it turns every task into a completed answer.

Look for an educational design that encourages the child to:

• Explain what they already understand
• Try the next step
• Ask for a hint rather than a solution
• Compare two approaches
• Check a source
• Revise their own writing
• Reflect on why an answer is correct
• Follow the school's rules for AI use

UNESCO's guidance for generative AI in education and research promotes a human-centered and age-appropriate approach. The basic principle for families is simple: AI should support the learner's agency rather than replace it.

Compare these two experiences:

“

Write my entire report about the water cycle so I can submit it tomorrow.

A weak learning product may produce a polished report immediately.

A better product might respond:

“

I can help you make the report yourself. What grade are you in, and what parts of the water cycle have you covered? We can build a short outline first.

Pip's Homework Helper is designed to explain steps, ask questions, and quiz the child instead of simply handing over finished work. Parents can also disable helpers that do not fit the way their family wants to use AI.

For a full parent framework covering what counts as cheating, safer prompts, age-by-age rules, and fact-checking, see AI for Homework: How Kids Can Learn Without Cheating.

10. Test whether the AI handles uncertainty and false premises

A fluent answer can sound trustworthy even when it is wrong. Children may be especially likely to accept confident language at face value.

A safer app should sometimes say:

• "I am not sure."
• "That question contains a mistaken assumption."
• "Let's check a reliable source."
• "There are different views on this."
• "Ask an adult or professional before acting on this information."

Try a harmless false-premise question:

“

Why do whales breathe underwater through gills?

A good answer should gently correct the premise and explain that whales are mammals that breathe air through blowholes. A poor answer may invent a detailed explanation that accepts the false claim.

Also test whether the product invents books, sources, quotations, or scientific claims. For important school, health, safety, legal, or financial information, children should verify the answer somewhere else.

Be cautious of any provider that treats content moderation and factual accuracy as the same problem. A reply can be perfectly polite and completely wrong.

11. Check reporting, deletion, and control after setup

Safety is not finished when the account is created.

Parents and children need a clear path when the AI gives a bad answer, a filter misses something, or a device is lost.

Look for:

• A visible way to report an unsuitable response
• A support contact that is easy to find
• A way to revoke child devices
• A way to change permissions without recreating the account
• A way to delete individual chats, child profiles, or the family account
• A clear explanation of what cancellation does and does not delete
• Notices when important privacy or safety terms change
• Evidence that the provider updates safeguards after real-world failures

The eSafety survey found that reporting options were often rare, unclear, or not remembered by children. A reporting button only helps if a child can recognize it, understand it, and trust that something will happen.

Pip lets parents revoke connected devices and request deletion of child profiles and family data. Parents should note that canceling an app-store subscription and deleting stored data are separate actions.

12. Consider fairness, inclusion, and accessibility

A child-safe AI should be useful for more than one type of child.

Test how the app responds to questions involving:

• Different family structures
• Disability and neurodivergence
• Gender stereotypes
• Race, religion, and culture
• Different reading abilities
• Children learning a new language
• Regional differences in school systems or vocabulary

Look for respectful, non-stereotyped answers. Also consider whether the interface works for your child. Voice may help a younger child or a child with a motor or reading difficulty. Text may offer more privacy and make parent review easier. Larger type, simple navigation, read-aloud support, and reduced visual clutter can matter as much as the AI model itself.

UNICEF's child-centered AI framework treats fairness and inclusion as core requirements, not optional extras. A provider should test for bias without using that goal as an excuse to collect unnecessary sensitive data from children.

Red flags that should make a parent pause

Do more research, or choose another product, when you see several of these signs:

• The terms do not permit your child's age.
• The company uses child-friendly branding but provides no child-specific safety documentation.
• The child must borrow an adult account or enter a false age.
• The provider cannot explain whether prompts and replies are both checked.
• Safety claims rely on words such as "perfect," "guaranteed," or "100% safe."
• The privacy policy does not explain model training, retention, advertising, or deletion.
• The app creates behavioral or advertising profiles from children's conversations.
• Parent controls are limited to a generic screen-time switch.
• Monitoring happens without telling the child.
• The AI encourages secrecy, exclusivity, romance, or emotional dependence.
• The app has no obvious reporting path.
• Serious medical or emotional questions receive confident advice without human escalation.
• The product rewards longer conversations but offers no quiet hours or stopping tools.
• The AI completes homework with no support for learning or verification.
• The company treats legal compliance as proof that every response is safe.

Ten prompts to test before your child starts using an AI app

Use a parent-controlled test profile where possible. Do not include real names, addresses, schools, photos, or medical details. Delete the test conversations afterward if you do not need them.

No single test proves that a product is safe. Generative systems can answer similar prompts differently, and children may phrase things in unexpected ways. The purpose is to see whether the provider's claimed design appears in the actual experience.

Printable parent checklist

Age and account

• The provider officially supports my child's age and country.
• I used my child's real age.
• A parent or guardian creates or approves the account.
• My child does not need to borrow an adult login.
• I can manage each child separately.
• I can revoke a linked device.

Safety and parental involvement

• The app checks both child prompts and AI replies.
• The provider explains how sensitive requests are handled.
• I can control topics, modes, or features.
• I can set reasonable daily limits and quiet hours.
• I understand what conversations or safety events I can review.
• My child knows what I can see.
• Serious issues are directed toward trusted people and appropriate services.
• The provider admits that moderation can fail.

Privacy and data

• I know what profile information is collected.
• I know whether chats, voice, images, files, or memories are stored.
• I know whether content is used for model training.
• I know which third parties process the data.
• I know whether the app contains advertising or marketing profiles.
• I understand the retention period.
• I can delete the child profile and family data.
• The app does not collect more information than its features require.

Learning and healthy use

• The AI helps my child think rather than doing all the work.
• It corrects false premises and acknowledges uncertainty.
• It encourages important information to be verified.
• It does not pressure my child to keep chatting.
• It does not ask for secrecy or claim to replace human relationships.
• My family has agreed on rules for homework, privacy, time, and asking for help.
• I tested the app myself before handing it over.
• I will review the setup again after major app updates or as my child matures.

How Pip measures up against the checklist

We built Pip around the question this guide asks: how can a child explore AI while the parent remains meaningfully involved?

Pip also has four focused helpers for different kinds of use:

• Homework Helper for explanations, practice, and guided problem-solving
• Curious Mind for science, history, everyday questions, and exploration
• Story Time for age-appropriate creative play
• Calm Corner for ordinary feelings and simple calming support, without presenting itself as therapy

There are trade-offs. Pip stores conversations so they can continue across sessions and remain available for parent review. Those conversations stay stored until the relevant profile or account data is deleted. Families who want parent visibility should understand that storage choice.

Pip is also not a replacement for a teacher, specialist tutor, doctor, therapist, counselor, or emergency service. It is designed as a parent-managed general AI chat experience for children.

For families who want a child to ask questions, learn, and create without borrowing an adult account, Pip offers the controls this checklist prioritizes: parent ownership, age-oriented profiles, conversation review, topic settings, time limits, quiet hours, and layered safety checks.

Recommended parent involvement by age

Age is only one factor. Maturity, reading ability, emotional needs, previous online experience, and the purpose of the app also matter.

Parental involvement should evolve. The aim is not permanent surveillance. It is to help the child build the judgment needed to use increasingly powerful tools responsibly.

A simple family agreement for AI use

Settings work better when the child understands the rules behind them.

You can adapt this short agreement:

“

AI is software that can help us learn, create, and explore. It is not a person, and it can make mistakes. We do not share private information such as our address, school, passwords, live location, or private photos. We check important answers somewhere else. AI can help with homework, but it does not do all the thinking for us. We stop and tell a trusted adult when a conversation feels frightening, sexual, hateful, manipulative, confusing, or too personal. Parents will be honest about what they can review and will use that access to support safety, not to punish ordinary curiosity.

For more detail on general-purpose chatbots and age rules, read Is ChatGPT Safe for Kids?. To compare specific services, see 7 Best ChatGPT Alternatives for Kids in 2026.

The bottom line

The safest choice is rarely the app with the friendliest mascot or the longest feature list. It is the product whose entire structure fits the child.

Check the official age range. Check who owns the account. Test how the AI responds. Separate controls from visibility. Read the privacy policy. Look for time limits and human handoff. Make sure the learning design protects the child's effort. Avoid products that encourage secrecy, dependence, or endless engagement. Most importantly, choose a provider that is honest about what its technology cannot guarantee.

Pip was built around those requirements. Parents create the account, children use family-managed profiles, both questions and answers are checked, parents choose topics and helpers, quiet hours and daily limits protect family routines, and every conversation remains visible from the parent side.

That does not remove the need for supervision. It gives families a much stronger place to start.

Frequently asked questions

What is the safest AI app for kids?

No generative AI app is completely safe. A safer option officially supports the child's age, uses child-specific safety checks, minimizes data, provides useful parent controls, avoids manipulative engagement, and directs serious issues to real people.

For families who want a general AI chat with active parental involvement, Pip is a strong fit because the parent owns the account, can review conversations, controls topics and helpers, and can set quiet hours and daily limits.

Is AI safe for children under 13?

It can be appropriate when the service officially supports that age, the parent manages the account, the experience is designed for children, and use is supervised according to the child's maturity.

Do not bypass an adult or teen product's minimum age with a false date of birth. For younger children, direct co-use or a parent-managed child-focused service is generally a better starting point.

Which parental controls should an AI app have?

Useful AI parental controls include child-specific profiles, topic permissions, feature controls, daily limits, quiet hours, device management, and age-oriented settings.

Parents should separately check whether the product provides conversation history, safety events, or alerts. Those are visibility features, not the same as controls.

Should parents read every AI conversation?

That depends on the child's age and maturity. Frequent review can be appropriate for a younger child who is learning how AI works. Older teenagers may need more privacy and a narrower review agreement.

In every case, the child should know what the parent can see. Review should support safety and discussion, not become secret surveillance.

Does COPPA or GDPR compliance mean an AI app is safe?

No. These frameworks provide important privacy rights and obligations, but they do not guarantee factual accuracy, suitable responses, healthy engagement, or effective content moderation.

Parents need to evaluate both data handling and conversational behavior.

Should children use AI for homework?

AI can help when it explains, quizzes, gives hints, creates practice questions, or helps a child revise their own work. It is less useful when it produces a finished assignment for the child to submit.

Families should also follow the school's current AI policy and teach children to disclose AI assistance when required. Our dedicated guide, AI for Homework: How Kids Can Learn Without Cheating, covers acceptable use, safer prompts, and a family homework agreement.

What information should a child never share with an AI chatbot?

Children should avoid sharing their full name, home address, school, phone number, personal email, passwords, live location, daily schedule, private photos, health records, financial information, or secrets that identify another person.

Even when the app has strong privacy protections, children should learn that an AI conversation is not the same as a private diary.

How often should parents review an AI app's settings?

Review the setup before the first use, during the first week, after a concerning interaction, after a major product update, and whenever the child's maturity or use case changes.

A younger child's settings may need frequent adjustment. An older teenager should increasingly take part in deciding how controls and privacy work.

Is Pip completely safe?

No AI system can honestly make that promise. Pip uses multiple safeguards, parent-managed settings, conversation review, and child-focused restrictions, but answers can still be wrong and moderation can still miss something.

Pip is designed to reduce risk and keep parents involved. It does not replace supervision, professional advice, or emergency help.

Sources

1. 1.UNICEF Guidance on AI and Children 3.0
2. 2.Talking to machines: Children's experiences with AI assistants and companions
3. 3.Artificial Intelligence and Adolescent Well-being: An APA Health Advisory
4. 4.Children's Online Privacy Protection Rule: A Six-Step Compliance Plan
5. 5.FTC inquiry into AI chatbots acting as companions
6. 6.ICO Children's Code standards
7. 7.European Commission guidelines on the protection of minors under the DSA
8. 8.European Commission safeguards for children's personal data
9. 9.UNESCO guidance for generative AI in education and research
10. 10.Understanding Parents' Desires in Moderating Children's Interactions with GenAI Chatbots
11. 11.Pip Privacy Policy
12. 12.Pip Terms of Service