AI Safety Methodology

• Encourage calmer, child-centred communication.
• Protect user privacy and data security.
• Produce clear, time-stamped records suitable for legal processes.

Safe messaging uses layered detection for insults, threats, harassment, and escalatory wording. Users keep full control to review or edit messages before sending. Models are tuned to avoid over-blocking while discouraging harmful language.

We log only the metadata required for functionality and export. Output is formatted for readability in PDF reports.

• Data is stored in reputable cloud infrastructure with strict access controls.
• Encryption protects data in transit and at rest where available.
• Internal tooling follows role-based access principles.

See the Privacy Policy for additional detail.

PDF exports include timestamps, clear sender/recipient labels, and pages designed for readability. Exports are intended to support legal processes—always seek legal advice for your jurisdiction.