Last updated
Last updated: June 22, 2026. Sprout AI is built to help children, parents, schools and trusted adults in a safe, calm and privacy-aware way. AI in Sprout Guard should support safety, not replace adults.
AI Safety Policy
Sprout Guard's AI safety policy explains how Sprout AI, AI Guard, Safe Writing, Parent Coach and AI Work Scanner may be used safely with privacy by design and backend-only AI.
What Sprout AI can help with
Sprout AI can explain why something is blocked, suggest safe replies, help a child pause before sending something, open relevant guides, explain red flags, support Help Me Now, give parents calm next steps, help teachers with classroom focus and explain rules in child-friendly language.
Sprout AI Guard
Sprout AI Guard can be active in enabled safety features such as Check Message, Scan Before Send, Safe Writing, Safe Browser, Safe Camera status, AppBlocker, Gaming, Gambling, Help Me Now, Parent Coach, Teacher Control and Guides. Sprout AI does not secretly read everything on the phone.
Safe Writing
Safe Writing helps a child avoid sending something that could harm others. This may include bullying, threats, image pressure, sexual pressure, sharing private information, hate, harmful messages, scams or gambling links. Sprout AI can suggest a calmer or safer reply.
Examples of safe replies
Sprout AI can suggest short replies such as: I need a break. Let's speak respectfully. I do not send images or codes. Children should be able to ask for help without shame or punishment.
Grooming and image pressure
Sprout AI can help explain red flags related to grooming, secrecy, image pressure, gift-card pressure and threats. In high-risk situations, the app can help the child contact a trusted adult with calm messages such as: You are not in trouble. Do not send anything yet. Sprout Guard can help you.
Parent Coach
Parent Coach helps parents understand an alert and respond calmly. It can suggest next steps such as: Start calmly. Say: Thank you for telling me. You are not in trouble. Parent Coach must not be used as proof of what happened.
AI in schools
Schools can use Sprout Guard for AI control during lessons and exams if consent is in place. A school can block external AI tools during exams, allow approved AI tools, use AI Work Scanner as a risk assessment and see limited metadata about AI access if consent allows it.
School limitations
Schools must not see private AI prompts, private messages, images, screen content, precise location or private family alerts. AI Work Scanner is a risk assessment, not proof.
AI Work Scanner
AI Work Scanner can help assess possible AI use in schoolwork, but results can be wrong. It should support good questions: Can the student explain the process? Are there notes or drafts? Which sources were used? Which part was assisted?
What Sprout AI must not do
Sprout AI must not help with grooming, manipulation, threats, harassment, bullying, image pressure, sexualized content, gambling tips, betting strategies, cheating on schoolwork, hiding traces, bypassing rules, pressuring others for images, money, gift cards or codes, self-harm or illegal behavior.
Privacy by design
Sprout Guard is built for minimal data by default. Safety events should store category, severity, timestamp, source and recommended action, not raw messages, full conversations, intimate images, screen data or unnecessary URL and search data.
Technical standards
Our default position is rawTextStored=false, rawOutputStored=false, fullUrlStored=false, searchQueryStored=false, imageStored=false for blocked content, imageSentToAI=false, screenCaptured=false, rawMessagesStored=false and screenshotsStored=false unless there is a clear, lawful and necessary purpose.
Backend-only AI
AI keys must not be placed in Flutter or frontend code. OpenAI, Anthropic or other AI providers may only be used through backend services with input sanitization, output guards, rate limits, logging without sensitive content, secure validation and fallback on errors.
Images and Safe Camera
Safe Camera must not send images to AI. Blocked private images must not be uploaded to parents, backend services or AI. If a local safety model is missing, Safe Camera must clearly show that setup is required, and the app must not pretend image scanning is active.
AI can be wrong
AI can misunderstand context. Sprout Guard should therefore use local rules first for high-risk cases, avoid automatic accusations, avoid automatic punishment, use AI as help and not proof, offer adult help and show calm explanations.
Reporting
If you discover a problem with Sprout AI, an unsafe response or an error in the safety system, contact security@sprout-guard.com. Do not include sensitive child data, private messages or images in the report.