Security

Security and Data Protection

This page explains in one place how adall protects your data. It answers the questions procurement, quality and information security teams usually ask; if something is missing, write to tech@adall.io and we will add it here. It is for information; the contractual terms are in the Terms of Use and the Data Processing Agreement.

In short

adall connects to your ad accounts only through official APIs (Google Ads API, Meta Marketing API and so on). It never sees your password and never uses screen scraping or browser automation.
The access keys the platforms issue are kept in an encrypted vault and are never written anywhere in plain text.
Each customer's data lives in its own workspace; automated tests check on every release that no customer can reach another's data.
Every change adall makes in your ad account is validated first, recorded, and can be undone; you set the spend limits.
Your data is never sold, never used for ad targeting and never used to train AI models.

Company

Pariette Teknoloji Ltd. Şti. (customers in Türkiye): Maslak Mah. Ahievran Cad. No: 11-2, Olive Plaza, 34398 Sarıyer / Istanbul, Türkiye. Maslak Tax Office, tax no. 7211005729. MERSİS 0721100572900001. Registered e-mail (KEP): pariette@hs06.kep.tr. Phone: +90 538 257 84 62. E-mail: tech@pariette.com.
Pariette Inc (customers in other countries): 8 The Green, Ste D, Dover, Kent County, DE 19901, USA. EIN 30-1424178. Phone: +1 302 754 3461.
Security and data protection contact: tech@adall.io
Agreements and approvals are made electronically only; no paper contracts with wet signatures are issued.

Where the data is

Customers in Türkiye: application and database in Türkiye, in CloudSpark data centres (Istanbul, Ankara, Izmir). These customers' data is processed and stored in Türkiye.
Customers in other countries: application and database at Hostinger, Boston (USA).
Backups and files: Cloudflare R2 object storage, encrypted.
Every subprocessor, its location and the data it handles: Subprocessors and Data Location.

What we process, and what we do not

Advertising data: campaign structure, settings, spend, impressions, clicks, conversions, search terms, location, device and hour reports, change history.
Sales data (if you choose): order and lead records your site sends: amount, currency, date, order number, ad click id.
Your customers' personal data: only if the buyer consented, and only as SHA-256 hashes (irreversible) of e-mail and phone. Hashing happens on your site; plain e-mail addresses or phone numbers never reach adall.
User account: name, e-mail, password hash, language, session and security records.
What we do not process: card details (they stay with the payment provider), ad platform passwords, your customers' plain contact details, special categories of personal data.

Encryption

In transit: every connection uses HTTPS (TLS); internal traffic between servers is not exposed to the internet.
Platform access keys: envelope-encrypted in a vault (XChaCha20-Poly1305). Each ciphertext is bound to its row, workspace and purpose, so a copy elsewhere does not decrypt. Vault keys exist only in the server environment and can be rotated. Every vault access is logged.
Short-lived access tokens are kept only in the memory of the running process, never in the database.
Passwords: stored only as one-way hashes.
Backups: the nightly database backup is encrypted with age; the decryption key is not kept on the server.

Access control

Roles: Owner, Admin, Analyst and Viewer. Each role can do only what it is permitted to; permissions are enforced on the server.
Two-step verification: available to every user, with recovery codes and lockout protection.
Sessions: expire after inactivity; all sessions can be signed out in one click.
Abuse protection: sign-in, sign-up and password reset are rate-limited; sign-up uses bot protection (Cloudflare Turnstile) and a risky-IP check.
Ad account permission: connections are read-only by default. Permission to make changes is granted in a separate, explicit step. When a connection is removed, its access key is revoked at the platform and deleted from the vault.

Access by adall staff

adall staff can see your data only through a time-limited, read-only support view that you approve, during a security investigation, or where the law requires.
Every access appears in your activity log.
Access to production servers is limited to the founding team; secrets live only in server environment variables, never in code or logs.

Isolation between customers

Every record belongs to a workspace and every query is limited to that workspace.
An automated test suite calls every API route with another customer's identifiers on every release to prove no data leaks. Code that fails it is not released.

Changes in your ad account

Every change is first tried in the platform's validation mode, then applied, then read back from the platform to confirm it.
Every change is written to the decision journal with its before and after state and can be undone.
You set the daily spend cap and how much adall may do on its own. The code enforces the limits; an AI suggestion never goes straight to your ad account, it passes the rules first.
Activity log entries cannot be edited or deleted.

Use of AI

adall's analysis and chat features use an AI provider (Anthropic Claude; Google Gemini for translation).
Only computed summary figures and the text a feature needs are sent to the model. Your customers' personal data and access keys are not.
Providers process the data under contract and only for this service; your data is not used to train models.
Code computes the numbers; the model only interprets them. Decisions pass the rules and your approval.

Backups and continuity

The database is backed up every night; the backup is encrypted and uploaded with a checksum to a different provider from production (Cloudflare R2), and kept for 35 days.
If a backup fails, or the last good one is older than 30 hours, the team is alerted at once.
Restores have been rehearsed; the drill is repeated every three months as the data grows.
Raw platform responses are archived separately, so reports can be rebuilt from the archive after an error.

Monitoring and incident response

Servers, queues, syncs, backups and error rates are monitored continuously; critical events alert the team immediately.
Service status is public: adall.io/status.
Breach notification: if we learn of a security breach affecting your data we tell you without undue delay and within 72 hours at the latest: what happened, which data was affected, what we did, what you need to do. We provide what you need as the data controller for your own notification to the authority.
If needed, platform access keys are revoked in bulk and reconnection is requested.

Secure development

Code changes are small and recorded; before release each one passes automated tests (more than 1,600), static analysis and format checks.
Architecture tests enforce that code writing to an ad account can be called only through one supervised gate.
Security design follows a threat model and an OWASP-based checklist; dependencies are kept up to date.
User-uploaded files are scanned for malware (ClamAV); if scanning is unavailable the upload is refused.
Logs never contain passwords, tokens or customer identifiers; user-supplied text is truncated when logged.

Certifications and independent audits

An independent audit for information security certification is in progress; the certificates will be published on this page once issued. The controls above are applied and continuously checked internally.
Google app verification: brand verification is complete; data access verification is under Google's review.
We answer enterprise customers' security questionnaires in writing.

Deletion and return of your data

You can export your data at any time; the download link is valid for 48 hours.
When you close a workspace you can reopen it during a 30-day grace period; after that the data is deleted from every store and a deletion record is kept.
Retention periods: Privacy Policy.