Data handling

Training, by plan

On Free, Student, and Professional, the CAD geometry, prompts, and metadata you submit may be used to operate, maintain, and improve CADABRA and its AI models. This is the default, and it is what the EULA and privacy policy describe.

On Enterprise, zero-retention handling is available under the agreement and is configured for your organization when that agreement is set up.

CADABRA does not sell your data to third parties on any plan.

Turning it off

One setting in Preferences stops both retention and training. With it on, your designs, prompts, and files are processed only to answer the request in front of them and are kept on no CADABRA server.

It is included with every plan.

• It also turns off cloud chat history and cross-device recall, because both of those are stored data by definition.
• It applies from the moment you turn it on. It is not retroactive, so if you want earlier content removed as well, ask us.

What applies on every plan

• Per-organization isolation. Indexes and processing are scoped to one organization and do not cross a tenant boundary.
• Encryption in transit and at rest.
• Minimum context. Only what a given prompt needs is loaded at run time, rather than your whole library.
• Never sold. CADABRA does not sell customer data to anyone.

What is kept even under zero retention

The privacy policy lists the categories that survive it:

• Operational diagnostics.
• Feedback you explicitly submit.
• Drawing-standard reference files, because a learned standard is a thing the product is designed to keep for you.
• Usage records used for billing and abuse prevention.

Your learned drawing standard and its reference files are accessible to authorised CADABRA staff under least-privilege controls, for support only.

Sub-processors and vendor review

The privacy policy names the third parties involved in operating the service and states what each of them receives. If you are running a vendor security review, read §3 first, then send us the questionnaire and we will answer it.

What has to leave the workstation

CADABRA needs connectivity, so it does not run fully offline.

Zero-retention handling and encryption at rest are the controls available for sensitive work. Model inference does require sending the prompt and the assembled model context off the workstation, so if your requirements rule that out entirely, tell us early and we will say plainly whether we can meet them.

Reporting a vulnerability

Send it to team@cadabrai.com. The same contact is published at /.well-known/security.txt.