Data processing — summary
Who is responsible for what
When your agency uses the software, you decide what happens to your clients’ data: you are the controller. We process that data only on your instructions, as your processor under Art. 28 GDPR: Ziad Amer – adbirdie, Lehrter Str. 24B, 10557 Berlin, Germany. The full data processing agreement is signed as part of onboarding.
We send the full agreement with the order form, and before it to anyone who asks.
What we process on your behalf
- The campaign, ad set, ad and performance data the software reads from the Meta ad accounts you connect.
- The Meta System User token you connect, stored encrypted.
- Your team members’ accounts: email addresses and sign-in credentials.
- Conversations with Birdie, the proposals it prepares, the approvals your team gives, and the audit record of each.
- Your clients’ decisions and comments on the sign-off pages you send them.
- The notification emails the software sends your team.
- Dictated speech, which is processed in memory for the length of one request and never stored.
- The briefs your team writes for image drafts, your clients’ brand files (logos, product photos, reference pictures and fonts) and the drafts made from them.
Where it is hosted
Everything above is stored in the European Union: the application, the database and the backups run with Render in Frankfurt, Germany, and the files move from Render to a Cloudflare bucket restricted to the EU. Speech is transcribed on our own servers in the same region. Image drafts are made by Google in its EU multi-region, which keeps the processing in the EU. Two providers in the table below process data outside the EU, under this agreement: Anthropic, whose AI model reads the account data in the USA, and Google Workspace, which sends the emails.
Who helps us
| Sub-processor | What for | Where |
|---|---|---|
| Render Services, Inc. | Hosting of the application, database, files and backups | EU (Frankfurt) |
| Anthropic Ireland, Limited | The AI model that reads account data and prepares proposals | USA (Anthropic, PBC), under the EU standard contractual clauses in Anthropic's data processing addendum |
| Google Cloud EMEA Limited (Workspace) | Sending the notification emails and the invitations | Google's data centres, including the USA, under the EU–U.S. Data Privacy Framework and the EU standard contractual clauses in Google's data processing terms |
| Cloudflare, Inc. (R2) | Storage of files: the creative library, chat files, brand files and image drafts | EU (a bucket restricted to the EU) |
| Google Cloud EMEA Limited (Gemini Enterprise Agent Platform) | Making image drafts from your team’s brief and your clients’ product photos and reference pictures; no training on them | EU (Google’s EU multi-region) |
Meta is not our sub-processor: you connect your own Business Manager, and your relationship with Meta stays yours.
How it is protected
- Meta tokens are encrypted by the application with keys held per workspace, never written into code, configuration or logs.
- Every workspace is separated from every other both in the application and in the database itself.
- Every proposal, approval and action is written to an append-only, hash-chained audit record, which the database itself refuses to change or delete.
- Every account signs in with a second factor.
How it is deleted
Image drafts your team does not approve are deleted after 30 days. When you leave, we hand you an export of your audit record first. Then your workspace’s data, your creative files and the stored token are deleted, within 30 days of the end of the agreement — all but the audit record.
The audit record stays: it is the record of what the software did in your ad accounts, the same one you received. It holds no free text; your team members’ email addresses on its entries are the one personal detail in it. How long it is kept is set in the full data processing agreement.