e.EverypathOpen composer

Everypath · AI and privacy

Useful assistance should never require blind trust.

Our current composer works locally without sending your draft to an AI model. Any future AI integration must be transparent, optional and independently reviewed before processing sensitive information.

Open the local-only composer →

What is available now

The public composer offers structured template drafting, editable output, local quality prompts, printing and text download. Everything typed in the composer stays in the current browser session. We do not intentionally set analytics cookies, transmit your entered text or store the draft in our application.

Your device, browser, extensions, print service, operating system and network may have their own security considerations. Use a trusted device and avoid entering personally identifying or special-category information in public demonstration tools.

What is not yet live

AI-generated drafting, automated case analysis, private records, secure file uploads, outbound messaging and shared professional workspaces are not available in this release. An optional invitation-only adult sign-in pilot may be activated separately; it is not a child-record service or an AI integration. We will not describe template generation as AI.

Explore the adult workspace preview →

How a future secure AI mode would work

  1. Explicit choice: the user decides whether to use AI for a specific action, sees what information will be sent, and can remain in local-only mode.
  2. Data minimisation: offer redaction and pseudonymous examples; do not assume automatic redaction catches every identifying detail.
  3. Identity and access: authenticated sessions, role-based access, rate limiting, session handling and audit controls that avoid copying sensitive prompts into logs.
  4. Contractual safeguards: verify model provider, region, processor terms, possible subprocessors, cross-region inference and retention; no unsupported claim of exclusive UK data residency.
  5. Human approval: AI suggests wording, questions and plans but never automatically contacts schools, makes decisions about a child or files legal documents.
  6. Safety boundaries: clear routes for safeguarding concerns, inaccurate outputs, discriminatory language, disputes and escalation to qualified humans.
  7. Ongoing assurance: security testing, data protection impact assessment, notices, incident response and regular independent review of model behaviour.

Special-category and children's data

Records about a child's disability, health or educational support can be highly sensitive. Before any AI use that processes such information, the operating organisation must identify the applicable lawful basis and any special-category condition, document the risks and complete appropriate privacy and safeguarding reviews.

Do not paste real child records into this public composer. Our current public tool does not use AI or a secure personal-record service.

Independent guidance

This describes planned safeguards, not an assertion that an AI system has passed approval. The operator/controller and full personal-data privacy notice are still subject to sign-off.