Trust Center

Version 1.0 · Effective 13 November 2026 · This is the first published version.

OliApp is operated by FreySoft Poland Sp. z o.o., a Polish company. Recruiters use OliApp to manage candidates, which means we hold personal data on behalf of our clients. This page sets out how we protect it, who else touches it, and what we can show you.

We have tried to make this page useful to a reviewer rather than reassuring to a reader.

Security questions: privacy@oliapp.co

1What OliApp is, in data-protection terms

Who we areFreySoft Poland Spółka z ograniczoną odpowiedzialnością, ul. Domaniewska 45, 02-672 Warszawa, Poland. KRS 0000864848, NIP 5213910266
Our roleProcessor for the candidate data our clients hold in the product. Controller for data about our clients' own users — accounts, billing, product activity, support
SurfacesThe Platform at app.oliapp.co, our website, the browser extension, the Slack integration where a client connects it, and the mobile application once released
What we processCVs and application materials, contact details, fields extracted or inferred from CVs, pipeline history, conversation summaries where a user has enabled message sync, and recruiter activity data
Supervisory authorityUrząd Ochrony Danych Osobowych (UODO), Poland

Full detail is in the Privacy Policy and the Data Processing Agreement.

2Where data is processed

The Platform itself — its databases, servers and file storage — runs on Google Cloud Platform, contracted through Google Cloud Poland Sp. z o.o., in European regions only. AI Search and Resume Breakdown run in the European Union too; features that run on a user's own AI provider key are covered below the table.

RegionWhat runs there
europe-central2 — Warsaw, PolandDatabase, application servers, backend services and functions, file storage
europe-west4 — Eemshaven, NetherlandsAdditional backend services and functions, file storage
europe-west3 — Frankfurt, GermanyFile storage
eu — European Union multi-regionAI Search and Resume Breakdown (Vertex AI)

Uploaded CVs and attachments are held in Google Cloud Storage in those regions and served directly from it. We use no content delivery network. AI Search and Resume Breakdown call Gemini models through Google Cloud's Vertex AI at its European Union endpoint, so the text of a CV or a search passed to them is processed within EU member states; only an AI Search request that carries a user's own AI provider key goes elsewhere (below).

The text sent to the models is not used to train or fine-tune them — that is a term of Google Cloud's service. Google may keep prompts for a limited period to detect abuse, and may cache inputs and outputs for up to 24 hours in the same region and for our project only, to answer faster.

Your own AI provider key. Conversation summaries (those made by message sync on an OpenAI key only), AI-drafted messages and template improvement run only on an OpenAI or Google Gemini API key that a user enters in their own settings, as does AI Search when a request carries such a key. That text goes to the provider the key belongs to, under the terms of the provider account the key belongs to, and may be processed outside the European Economic Area. The provider is the organization's processor, not our sub-processor; see Privacy Policy §11.1.

We store no candidate data outside the European Economic Area; text sent under a user's own AI provider key goes, on the organization's instruction, to the organization's own provider, as described above. Where a provider's non-EEA parent can reach data for support, that access is treated as a transfer and is covered by the mechanism named on the Sub-processors page.

Transfers outside the European Economic Area, and the mechanism relied on for each recipient, are listed on the Sub-processors page.

3Sub-processors

We publish the full list, with the jurisdiction, the purpose and the transfer mechanism for each: Sub-processors.

Integrations a client connects themselves — the Slack integration is the current example — are not our sub-processors. Data reaches them on the client's instruction, into a workspace the client owns and contracts for. See Data Processing Agreement §8.

The Slack app requests bot-level permissions only — links:read, links:write, chat:write, commands — and does not read channel message history. Workspace and user tokens are held in a VPC-isolated database in our European infrastructure, encrypted at rest; channel identifiers are not stored persistently; transient payloads are purged within 24 hours. Uninstalling revokes and deletes the tokens immediately.

You can subscribe on that page to be notified of changes. Under our Data Processing Agreement we notify clients at least 30 days before a new sub-processor is added, and you may object within that period.

4Security measures

Stated as at the date at the top of this page.

4.1 Encryption

At rest. AES-256, applied by the underlying Google Cloud infrastructure to compute disks, object storage and the database backups. Keys are managed by Google.

A second layer is applied at application level (AES-256-CBC) to the fields where exposure would hurt most: integration API keys, Slack installation data, the support-chat token, candidate data aggregates, the audit trail of operations on candidate data, conversation texts and summaries, and CV upload-link secrets.

In transit. TLS 1.2 is the minimum on every public endpoint, with no fallback to earlier versions; plain HTTP is redirected to HTTPS. The identity provider requires SSL for all external requests. Connections between services inside our private network are restricted by firewall rules to the specific services that need them, and the internal parsing services are not reachable from the internet at all.

4.2 Access control

Access to the production environment is held by five named members of staff: two with full administrative rights, two with server and database access, and one with read access to logs only.

Server access is never by exposed credentials. It runs through an identity-aware tunnel tied to a Google account, so there is no SSH port open to the internet and no shared key to leak.

Multi-factor authentication is enforced at organisation level for both the cloud console and source control. It cannot be switched off per account.

Deployment runs through workload identity federation: the CI system holds no long-lived cloud credentials. The one long-lived service-account key that exists belongs to the monitoring integration and carries read-only monitoring permissions, with no access to personal data.

4.3 Logging and monitoring

We log application events, web access, administrative actions in the cloud, administrative actions inside the product, and user sign-in events.

Personal data and tokens are redacted before logs leave the application — this is applied at the logging layer itself, on every channel, rather than left to each call site.

Retention: 7 days on the server, 30 days in the cloud logging store, 400 days for cloud administrative-activity records, and 365 days for an archive of error-level logs held in europe-central2.

Alerting is automatic. Rules are kept as code and cover service availability, host resources, database connections, server errors, missing telemetry and missed backups. Alerts go to an on-call channel and have a written runbook.

4.4 Backup and recovery

Databases are backed up daily at 02:17 UTC to object storage in europe-central2, separate from the production machine. Dumps are retained 30 days, with a seven-day window during which an object cannot be deleted and a seven-day soft-delete window on top.

A "backup missing" alert fires if no successful upload has completed in 26 hours, so a silent failure surfaces the next morning rather than at the moment it is needed.

4.5 Incident response

Security incidents are handled under FreySoft's information-security incident procedure, which sets out roles, classification, and the stages containment, investigation, recovery and closure.

Operational alerts have a named on-call owner and a runbook.

Our contractual commitment to notify an affected client within 48 hours of confirming an incident affects their data is in Data Processing Agreement §5.4.

To report a suspected incident: privacy@oliapp.co

4.6 Vulnerability management

Container images and repositories are scanned with Trivy for known vulnerabilities and for committed secrets. Scans run on pull requests that change dependencies or container definitions, and on a recurring schedule. A build fails on a finding, and the list of accepted findings can only shrink — an existing exception may be removed, but new ones are not added silently.

The backend's PHP dependencies are audited on every pull request against the same ratchet.

5Breach notification

Where we confirm a personal data breach affecting customer data, we notify the affected customer without undue delay and in any event within 48 hours of becoming aware of it. The notification states the nature of the breach, the categories of data affected, and the mitigation taken or planned.

The contractual commitment is in Data Processing Agreement §5.4.

6Data subject requests

Clients handle requests from their own candidates using the controls in the product; our role is to assist, as described in Data Processing Agreement §7.

Where a request reaches us directly rather than a recruiter, we can act across the whole platform: locate every record holding that person, export it, and erase it everywhere. Individuals can start that here: Your Data Request.

7Documentation

DocumentAvailability
Data Processing Agreementdownload — public, no request needed
Privacy Policy, Cookies Policypublic
Sub-processor listpublic
Completed security questionnaire (CAIQ or your own format)on request

To request a document, write to privacy@oliapp.co and tell us which one and why. We respond within five working days.

8Documents