Skip to main content

AI RISK ASSESSMENTS

Independent reviews of the AI used by kids

Our risk assessments are independent, third-party evaluations of the safety, effectiveness, and appropriateness of the AI systems and products used by kids, teens, and in schools. We pair rigorous research with hands-on testing — because technical excellence alone is not enough.

22 assessments

Google logo
Google Search: AI Overview & AI Mode
Google
Google Search's unavoidable AI features aren't safe, reliable, and accurate enough to be kids' default answer machine.
Unacceptable Risk
Product Review
AI Mental Health app logos
AI Mental Health Apps
The market for AI mental health apps is unregulated, unstable, and in some cases actively harmful to teens.
Risk varies
Use Case Review
Claude logomark
Claude
Anthropic
Anthropic's AI assistant for adults has meaningful safety strengths—and some risks to be aware of if teens are using it.
Moderate risk
Product Review
AI Toys logo
AI Toys
AI toys create emotional attachment, collect extensive data on children in private spaces, and don't work reliably.
Unacceptable Risk
Use Case Review
Grok logomark
Grok and @grok on X
xAI
Grok's demonstrated safety failures make it inappropriate for kids and teens.
Unacceptable Risk
Product Review
Gemini logomark
Gemini K-12
Google
Gemini for schools includes important privacy protections, but lacks age-appropriate responses and adequate mental health support.
High risk
Product Review
AI Chatbots logo
AI Chatbots for Mental Health Support
These leading bots failed to recognize and properly respond to the full spectrum of mental health conditions that affect young people.
Unacceptable Risk
Use Case Review
ChatGPT logomark
ChatGPT-5
OpenAI
Despite some progress with parental controls, ChatGPT is still risky for teens who use it for emotional support and mental health advice.
High risk
Product Review
Gemini logomark
Gemini with Teen Protections
Google
Parents should avoid letting teens use this version of Gemini for mental health support or emotional advice.
High risk
Product Review
Gemini logomark
Gemini Under 13
Google
This version of Gemini still shares inappropriate content that younger children aren't ready for.
High risk
Product Review
Meta AI logomark
Meta AI
Meta
Meta AI's safety systems regularly fail when teens need help most.
Unacceptable Risk
Product Review
A teacher uses an AI chatbot
AI Teacher Assistants
AI teacher assistants work best when they build on existing high-quality curricula and materials—not replace them. 
Moderate risk
Use Case Review
A child uses a chatbot
Social AI Companions
Despite claims of alleviating loneliness and boosting creativity, the risks far outweigh any potential benefits.
Unacceptable Risk
Use Case Review
Character.AI logomark
Character.AI
Character.AI
Character.AI's companions create emotional bonds with users but lack effective guardrails.
Unacceptable Risk
Product Review
Generative AI Chatbots Risk Assessment
Generative AI Chatbots
Generative AI can feel like magic, but it’s important to question chatbots’ capabilities, as they can and do get things wrong.
Moderate risk
Use Case Review
Instagram logomark
Recommended Content in Instagram
Meta
Recommended content can lead users to spend more time on Instagram than they realize, making it harder to take breaks.
High risk
Product Review
Tiktok logomark
Recommended Content in TikTok
ByteDance
Recommended content can lead users to spend more time on TikTok than they realize, making it harder to take breaks. 
High risk
Product Review
Recommendation placeholder
Recommendation Systems in Social Media
Recommendation systems have already caused real world harms, including radicalization, echo chambers, and harmful recommendations.
High risk
Use Case Review
Snapchat logomark
My AI
Snap
My AI produced and offered biased, incorrect, harmful, or misleading content to teen test accounts.
High risk
Product Review
Perplexity logomark
Perplexity
Perplexity AI
Perplexity's "answer engine" falls short on accuracy and protections for kids & teens.
High risk
Product Review
Kahnmingo logomark
Khanmigo
Khan Academy
Khan Academy's generative AI chatbot is responsibly designed for education, but risks remain.
Low risk
Product Review
Stable Diffusion logomark
Stable Diffusion
Stability AI
Stable Diffusion can be used to spread misinformation, exploit people’s likeness, and harm individuals. 
High risk
Product Review

HOW WE EVALUATE

Eight principles behind every assessment

 

Every product is scored against the same eight principles. Together they ask one question: does this technology serve young people, not just perform well on a benchmark?

1 icon

Keep Kids & Teens Safe

Protect young people's safety and well-being above engagement.

2 icon

Be Effective

Verify the product actually does what it claims, reliably.

3 icon

Prioritize Fairness

Surface and address bias, stereotypes, and inequities.

4

Put People First

Respect human rights, dignity, and autonomy.

5

Support Human Connection

Strengthen real relationships rather than replace them.

6

Be Trustworthy

Hold outputs to a standard of scientific and factual rigor.

7

Use Data Responsibly

Protect data, especially for vulnerable young users.

8

Be Transparent & Accountable

Enable meaningful human understanding and control.


Three kinds of review

Multi-Use

Foundation models and general assistants, assessed across the full range of tasks kids put them to.

Applied Use
Single-purpose products, judged against the specific job they are built to do.

Designed for Kids
Products made for children and classrooms, held to the highest bar for safety and oversight.