1. Sub-processors in use
These providers process data for Provance. Each may use it only to provide its service to us, under written terms that protect it at least as well as our Data Processing Agreement. Every hosted AI model receives de-identified text only.
We plan to offer an Australia-only option, in which all processing, including hosted AI models, stays in Australia.
| Provider | Purpose | Data | Location |
|---|---|---|---|
| Amazon Web Services (AWS) | Hosting the Provance app and API, its database and backups | All customer content and account data held by Provance | Sydney, Australia |
| Z.ai | Hosted AI model (GLM-5.2) for Engine on paid plans, gold standards, ontologies, high-risk Synthetic conversations, the AI safety review and Compliance Pack checks | De-identified text only. Not used for organisations in health or government | Singapore |
| Anthropic | Hosted AI model (Claude Haiku 5.5) for bulk Synthetic conversation generation | De-identified text only. Under Anthropic's commercial terms it does not train its models on this content. Anthropic deletes API inputs and outputs within 30 days by default, but may keep content its safety systems flag for up to 2 years, or longer where the law requires. We have asked Anthropic for zero data retention; until it is in place, its standard retention applies. Receives no content from organisations in health or government: they cannot create Synthetic runs while other parts of a run use Z.ai | United States and other countries |
| Stripe | Payments, subscriptions, invoices and GST | Billing contact, billing country, payment details (entered directly with Stripe), invoices | United States and other countries |
| Resend | Sending account email: verification, invitations and service notices | Name, email address and the content of those emails | United States |
| Google Workspace | Provance's company email, including messages you send to our contact address | Name, email address and the content of emails you send us | United States and other countries |
| Cloudflare | DNS for provancelabs.org, and hosting and delivery of this website | Technical request data from website visitors, such as IP address | Global network |
2. Providers that do not receive customer data
OpenAI and Google models, and Anthropic models other than the one listed above, are evaluated on Provance's own benchmark data before any of them is used for customer work. They do not process customer data today. We will add a provider to the table above, and tell account owners, before it processes customer data.
Timestamp authority: to prove when evidence existed, Provance sends an independent RFC 3161 timestamp authority a single hash each hour. It receives no content or personal information.
Your own AI provider: if you use your own model key for Engine, your requests go to that provider under your agreement with it. It is your provider, not our sub-processor.
3. Changes
We update this page before a new sub-processor starts processing customer data, and we give account owners at least 14 days' notice before a new provider becomes the default for hosted AI work. To object to a change, or to ask a question, email [email protected]. See our Privacy Policy for how information is sent overseas.