How we use AI

Last updated: 25 June 2026

We believe in being clear about where AI is involved. This page explains how Harbourstone uses Anthropic’s Claude API, what your data does and doesn’t do, and the limits of AI output.

1. Where AI is used

Harbourstone uses Anthropic’s Claude API for one thing: regenerating a dashboard. When you ask, Harbourstone sends the dashboard’s source and your feedback/annotations to Claude, which returns an updated version of the HTML. Hosting, access control, theming, text edits and versioning do not use AI.

2. Opt-in and human-in-the-loop

AI regeneration is opt-in per dashboard and limited to owners/administrators. A person starts each regeneration, reviews the new version, and chooses whether to keep it — and can roll back at any time. AI never publishes on its own.

3. AI-assisted, and labelled

Versions produced with AI are marked as AI-assisted so it’s always clear which content was generated with Claude’s help and reviewed by a person.

4. Limits of AI output

AI outputs may be inaccurate, incomplete or out of date. Review them before relying on or sharing them.
Harbourstone is a reporting and visualisation tool. AI outputs are not professional advice (financial, legal, medical, employment) and must not be used to automate decisions about people without human review and disclosure.

5. What happens to your data

Data is sent to Claude only when you trigger a regeneration — nothing is sent in the background.
What is sent: the dashboard source and your feedback.
Under Anthropic’s API terms, inputs and outputs are not used to train Anthropic’s models. Anthropic acts as our subprocessor (see our Privacy Policy).
On self-hosted deployments, data stays on your infrastructure apart from these explicit API calls.

6. Responsible use

Your use of AI features must follow our Acceptable Use Policy and Anthropic’s Usage Policy. We design the feature to keep a human in control and to avoid high-risk automated decision-making.

Questions: ai@harbourstone.kz.