AI Data Use

Training of AI models

ANCHOR does not use customer governance records, learning records, Trust Pack materials, clinic metadata, or account data to train public general-purpose AI models.

Current generation posture

• Live Workspace generation is production-off.
• The current live Workspace generation path has been built against Anthropic, but it remains production-off.
• Anthropic becomes a subprocessor when live generation is enabled.
• ANCHOR is presented as deterministic governed generation today and is architected for vendor-neutrality / vendor-neutral over time; present-tense vendor-neutrality is not claimed while a single provider is wired.

What this page does not claim

This page does not claim that no AI provider ever sees data, and it does not claim that data never leaves the UK. Those positions depend on technical and contractual arrangements that must be confirmed, including in the subprocessor and data-processing documentation prepared for solicitor review, before any paid pilot or real clinic data.

Customer data submission

Customers must not upload raw clinical records or identifiable client or patient material unless a specific authorised product flow and a corresponding agreement permit it. ANCHOR is metadata-only by default and is not intended to receive raw clinical content.

Status

Live generation remains production-off until the local/staging safety gate passes, and legal and subprocessor documentation must be complete before any paid pilot or real clinic data.