Safe Evidence CompanionA Safe Harbor child-safety projectCONCEPT PROTOTYPE

AI Safety Coach demonstration

Practice recognizing online pressure before it happens.

Short, game-like lessons could teach young people how manipulation works and let them practice a safer next step using reviewed, entirely fictional situations.

Fictional practice · Level oneDEMONSTRATION

“You won—but send your login code in two minutes.”

The safer move is to pause, share nothing, and check the official app. Urgency plus a request for a secret is a reason to stop and verify another way.

Notice

Look for urgency, secrecy, threats, and requests for private information.

Decide

Pause, protect private information, and verify through an official route you already trust.

Learn

See a plain-language explanation that can be used in a different situation.

Practice topicsPLANNED

Six kinds of online pressure

Each lesson would use a fictional, pre-reviewed situation and explain why one response is safer.

Suspicious messages

Notice pressure, secrecy, threats, and requests for private information.

Fake profiles

Check identity without sharing more information or confronting the account.

AI voice scams

Treat a familiar-sounding voice as a clue—not proof of identity.

Manipulated images

Check the original source, context, cropping, and independent reporting.

Phishing

Avoid unknown links and verify requests through an official route.

Social engineering

Practice pausing, leaving, setting a boundary, or asking for help.

Privacy by designREQUIRED

Practice would not require a child’s real messages.

The planned lessons use fictional cases. They do not need access to accounts, contacts, conversations, photos, microphone, camera, or location.

Current readinessHUMAN USE BLOCKED

Software mechanics passed. Human use remains gated.

The local development work reports 720 deterministic fictional cases and 21 of 21 mechanics tests passed. Those results do not establish developmental suitability, safety effectiveness, or public readiness.

Still required

Independent child-safety, developmental, accessibility, privacy, ethics, and reproducibility review.

Not yet allowed

No public lessons, child participation, real-message ingestion, automatic reporting, or safety claim.

KEEP EXPLORING