The Data Promise

How Harbour handles your organization's information — in plain language.

1. Your conversations are not used to train AI models

Harbour is powered by Anthropic's Claude models. By default, Anthropic does not use content submitted through its API to train its models. This is a standing commitment in Anthropic's commercial terms, not a setting we have to remember to turn on.

References: Anthropic Commercial Terms of Service · Anthropic Privacy Center

2. Your workspace is yours alone

Every document, conversation, and setting in this workspace belongs to your organization and is isolated to your organization's workspace. Every request in the product is scoped to your organization — other organizations cannot see your information, and you cannot see theirs.

3. Answers are grounded in your documents

When your documents cover a topic, the assistant grounds its answer in them and names its sources. When it answers from general knowledge instead, it says so — plainly, in the same place.

4. You control who has access

Access is invite-only, managed by your administrators. Sign-in uses your Google account — no extra passwords to manage. Administrators can disable an account at any time, effective immediately.

5. Your data is exportable and deletable

On request, we will export your organization's data — documents, conversations, settings — in a standard format, or permanently delete all of it. Leaving is as easy as joining.

6. Good practice still applies

We encourage your team not to paste personally identifying supporter information into any AI tool, including this one. The assistant works from your policies and procedures — it rarely needs personal details to be useful.