AI & Your Data

Your data is never used to train AI models. Not by ORCA, and not by our AI provider. This page explains exactly how AI works inside ORCA, what leaves the platform, what doesn't, and where you can verify every claim for yourself.

The short version

• ORCA's AI features are powered by Claude, built by Anthropic, and accessed through Anthropic's commercial API.
• Anthropic does not train its models on data sent through the API. This is a contractual commitment in their commercial terms, not a setting we had to switch off.
• Nothing is kept. Anthropic automatically deletes API inputs and outputs from its systems within 30 days. ORCA has not opted in to any data-sharing or development partner programme, so nothing is retained for model improvement.
• Your database never leaves ORCA. The AI does not have access to your database and no copy of it exists anywhere else. Each AI request sends only the small piece of context needed to answer that one question.
• AI in ORCA is read-only. It can summarise and answer questions about your records. It cannot create, edit, or delete anything.

Where your data lives

Your ORCA data lives in ORCA's own database, on infrastructure we control. AI providers are not part of ORCA's storage, backup, or hosting arrangements in any way. If every AI feature were switched off tomorrow, your data would sit exactly where it sits today.

What actually happens when you use an AI feature

When you ask ORCA AI a question, here is the complete journey your data takes:

1. Your question, plus the minimum relevant excerpt from your account (for example, a company summary or a handful of recent communications), is sent to Anthropic's API over an encrypted connection.
2. Claude generates a response, which is returned to you inside ORCA.
3. Anthropic deletes the request and response from its systems within 30 days. The only exceptions are where retention is required by law, or where content is flagged as violating Anthropic's usage policy.
4. ORCA logs the volume of usage (token counts) for billing purposes. The content of your questions and the AI's answers is not stored in those logs.

Every request is scoped to your account and to the permissions of the user asking. The AI can only see what that user is already allowed to see.

What never happens

• Your database is never bulk-exported, synced, or copied to any AI company.
• No AI model is trained, fine-tuned, or "improved" on your data, by us or by Anthropic.
• Nothing is shared between ORCA accounts. Your data is invisible to every other organisation on the platform.
• ORCA never submits your content to Anthropic's optional feedback or development programmes.
• The AI cannot modify your data. All AI access is read-only, enforced at the code level.

Which ORCA features use AI

AI is only involved when one of these features is actively used:

• AI Assist, the in-app assistant panel
• Reports: "Describe it" mode, which turns a plain-English request into a report
• How do I..., the in-app help tool
• Daily Briefings, the scheduled email digest
• Automation email drafting, where an automation includes an AI-drafted message

Browsing, editing records, searching, importing, and everything else in ORCA involves no AI processing at all.

Check it for yourself

You don't have to take our word for any of this. These are Anthropic's own published policies:

• Is my data used for model training? confirms API data is not used for training by default.
• How long do you store my organisation's data? confirms the 30-day deletion window for API inputs and outputs.
• Commercial Terms of Service is the contract governing ORCA's API usage.
• Data Processing Addendum sets out Anthropic's role as a data processor. Your organisation remains the controller of your data.
• Anthropic Trust Center covers Anthropic's security certifications and compliance posture.

Still have questions?

We'd rather you ask than worry. Contact us at hugh@dunkdigital.com and we'll answer anything this page doesn't cover.