Data Processing Agreement

1. Roles & scope

This Data Processing Agreement overview applies between TraceHold B.V. ("Tracehold", "we", "us"), formerly Verifia, established in the European Union at Werfkade 25, 1033 TA Amsterdam, Netherlands (KvK 97572896, VAT NL868118278B01), and the organisation that subscribes to a managed Tracehold plan ("you", the "Customer").

For managed deployments (the paid Insight and Governance plans):

• the Customer is the data controller: you decide why and how personal data is processed through your deployment;
• Tracehold is the data processor: we process personal data only on your documented instructions, as set out in the DPA and the configuration of your deployment. This includes the accounts of the people who use your console.

This page does not cover the free browser extension or this website. For those, Tracehold is the controller; see the Privacy Policy.

2. Subject matter & duration

The subject matter of the processing is the provision of the Tracehold managed AI-DLP service: detecting sensitive data before it reaches AI tools, recording an audit trail of findings, and producing the compliance reports and regulation mapping (GDPR, EU AI Act, DORA, NIS2, ISO 27001 and SOC 2) that the service provides.

The duration of the processing matches the term of your subscription. Unless the DPA states otherwise, Tracehold applies these platform-wide retention defaults, which are the same for all organisations: event records 365 days; masked samples and any kept content 90 days; endpoint-agent telemetry 90 days; browser-installation records 180 days after their last activity; console activity log, including the IP address, 30 days (older entries are removed whenever the service starts). These are the automatic purges: other derived records, such as the history of risk scores and incident reports, are not purged automatically and are deleted under the end-of-contract rule or on the organisation's request. At the end of the contract, data is returned or deleted as described in section 10.

3. Nature & purpose of processing

The nature of the processing is automated detection in the browser or on the endpoint, followed by the transmission of event data about findings to your Tracehold backend: type of data, severity, destination and action, plus a masked sample of the value. Prompt text reaches Tracehold only through the optional features described in section 4.

The purpose is strictly to deliver the service you have subscribed to:

• to detect and prevent sensitive data from reaching AI tools (across more than 300 AI tool domains, using more than 1,600 detection rules);
• to give your security and compliance teams an audit trail of findings; and
• to produce the compliance reports and regulation mapping covering GDPR, EU AI Act, DORA, NIS2, ISO 27001 and SOC 2.

We do not use the personal data we process on your behalf for any other purpose, we do not sell it, and we do not use it to train AI or machine-learning models.

4. Categories of personal data & data subjects

Categories of personal data

By default, the personal data we process as your processor is:

• Console users: name, work email, role, department, sign-in times, IP address and browser of console activity, and authentication data (password hash; second-factor secrets stored encrypted), directory group memberships and single sign-on identifier where those are used, language and time zone;
• Monitored users and devices: user identifier and work email, operating-system user name, computer name, directory domain, installation and device identifiers, browser and version and, for endpoint agents, the device's IP address and process names;
• Event data: time, AI tool and page address, action, severity, type of data detected, a masked sample of the detected value (values of up to 12 characters are masked completely; longer values show the first 8 and last 4 characters), for endpoint agents an excerpt of the text with sensitive values masked, and false-positive reports filed by users (masked fragment up to 500 characters and an optional note), kept 90 days;
• Derived data: risk scores per user and per organisation.

Optional, off by default. Tracehold never switches one on on its own initiative: they are switched on at your request, with your data protection impact assessment confirmation, by your administrators or by Tracehold staff acting on your documented instruction, and every change is recorded in the activity log:

• LLM classifier (Governance plan, beta): the text, the recent context of the conversation, older messages sent to be summarised and, where endpoint interception is enabled, text extracted from files, sent in transit; Tracehold does not store it;
• kept content of findings (the original text around a finding, fragments of conversations, requests or responses), encrypted at the application level and deleted after 90 days, with a masked version stored alongside it;
• files sent to the Tracehold backend by file fallbacks, read and not stored;
• the list of other browser extensions installed in monitored browsers (third-party extension management).

Detection looks for special categories of data (for example health data) as part of its purpose; by default their content is not stored.

Categories of data subjects

The data subjects are the Customer's staff: your employees, contractors and other authorised users, both monitored users and console users. If you enable kept content, third parties mentioned by your staff may also be involved; minimising this is within your control as the controller.

Your responsibilities as controller

These features are decided by you, as controller; Tracehold only switches them on on your documented instruction. You are responsible for having a legal basis for the monitoring, for carrying out any data-protection impact assessment it requires, and for informing your staff and, where required, your works council, including about the risk scores the console derives per user and per organisation.

5. Processor obligations & staff access

As your processor, and consistent with Article 28 GDPR, Tracehold commits to:

• process only on your documented instructions, including for international transfers, unless required to do otherwise by EU or Member State law (in which case we will inform you, where legally permitted);
• implement and maintain appropriate technical and organisational security measures (see section 7);
• engage sub-processors only under the conditions in section 6;
• assist you with data-subject requests and with your obligations on security, breach notification and data-protection impact assessments (see section 9);
• return or delete personal data at the end of the service, at your choice (see section 10);
• make available the information needed to demonstrate compliance and allow for and contribute to audits (see section 9); and
• notify you without undue delay, and in any event no later than 48 hours after becoming aware of a personal-data breach affecting your data.

Tracehold staff may access a customer's environment for support and operations through impersonation, an internal read-only analysis tool and platform administrator accounts, using individual named accounts. When acting through impersonation, staff hold the organisation's administrator role, including the permission to reveal kept content. Changes and reveals made through impersonation are recorded with the staff member's identity in the activity log, which is kept for 30 days; queries made with the internal analysis tool are recorded in a separate audit log. Kept raw content stays encrypted at rest and is revealed only through the reveal permission, which is itself logged.

6. Sub-processors

Current sub-processors: Contabo GmbH (hosting of the platform, including the database and Tracehold's own mail server), and, only when the LLM classifier has been switched on for your organisation, OVHcloud (AI inference; text in transit). The full list, with purposes, data and locations, is on our sub-processors page. The AI tools your organisation monitors and the destinations it configures (SIEM, Slack or Teams, identity providers) are your choices and are not Tracehold sub-processors.

We notify you of any intended addition or replacement of a sub-processor at least 30 days in advance, so that you can object. System and transactional email is sent from Tracehold's own mail server in the EU, so no third-party email provider acts as a sub-processor; we use no third-party error-monitoring or analytics service. The bot protection of the contact form on our website is described on the sub-processors page; it concerns website visitors, not data processed for you.

7. Security measures

Security is built into the product. The technical and organisational measures appropriate to this processing include:

Encryption

Data travels over HTTPS. The raw content of findings that your organisation chooses to keep, the stored secrets of LDAP and OIDC single sign-on and of SIEM integrations, and the secrets of authenticator apps (TOTP) are encrypted at rest with AES-256-GCM at the application level. SIEM secrets and webhook signing secrets are never returned by the API once saved. Masked excerpts, outbound-webhook signing secrets and the destination addresses of Slack, Teams and email alerts are stored without application-level encryption.

Isolation

Each organisation's data is separated by organisation: the application filters by organisation and, for requests from signed-in users and API keys, PostgreSQL enforces it again with Row-Level Security. Ingestion of endpoint-agent telemetry and sign-in run with explicit organisation filters in the application. Parent organisations in a multi-tenant space (Governance plan) can read their subsidiaries' data by design; Tracehold staff access is described in section 5.

Audited reveal

Where you keep the content of findings, the unmasked original stays encrypted until an authorised admin reveals it. Each reveal needs a specific permission and is logged: who looked, when, at what.

Detection rules delivery

The detection-rules manifest is signed with Ed25519 and delivered over HTTPS to authenticated clients. The extension verifies the signature against a public key built into it and the integrity of the rules (SHA-256) before applying them, rejects a rule set older than the last one it saw and, if a check fails, keeps the rules it already has.

Identity & access

Sign-in is by OIDC, LDAP/AD or password, with sign-in throttling and an optional second factor (authenticator app, email code, backup codes) that your organisation can require. Internal access is audited.

On certifications, we describe our status plainly: Tracehold does not currently hold ISO 27001, SOC 2 or any equivalent certification; its reports and evidence are mapped to ISO 27001 and SOC 2 controls to support your audits, and are not legal advice or a certification. You can read more on our Security page.

8. International transfers

Tracehold's production service is hosted in the European Union by Contabo GmbH. If the optional LLM classifier has been switched on for your organisation, the text is sent in transit to OVHcloud AI Endpoints, which OVHcloud states are hosted in France. Processing of your data under this DPA otherwise takes place in the EU.

If a transfer outside the EU/EEA were ever required for your deployment, it would happen only on your documented instructions and under an appropriate transfer mechanism, such as the European Commission's Standard Contractual Clauses, with supplementary measures as needed. The website contact form uses Cloudflare Turnstile (Cloudflare, Inc., United States); that concerns website visitors and is described in the Privacy Policy.

9. Audits & assistance

We make available the information reasonably necessary to demonstrate compliance with our Article 28 obligations, and we allow for and contribute to audits, including inspections, conducted by you or an auditor you mandate, on at least 30 days' prior written notice and no more than once per year (or following a substantiated personal-data breach), subject to confidentiality and to not compromising the security or isolation of other customers.

Taking into account the nature of the processing and the information available to us as a processor, we will assist you:

• in responding to data-subject requests (access, rectification, erasure, restriction, objection, portability) relating to data in your deployment;
• with your obligations on security of processing, breach notification and communication to data subjects; and
• with data-protection impact assessments and any prior consultation with a supervisory authority.

10. Data return & deletion

On termination or expiry of the managed service, and at your choice, we will return the personal data we process on your behalf, or delete it, and delete existing copies, unless EU or Member State law requires us to retain it.

We do this at your request, within a maximum of 30 days of the request. Until the data is deleted, the retention defaults in section 2 continue to apply.

11. Liability

The allocation of liability between controller and processor, including any caps, exclusions and the interaction with the liability terms of your main agreement, is governed by the liability provisions of the EULA, the signed order and the Data Processing Agreement, following the apportionment principles of Article 82 GDPR. This overview page creates no contractual liability on its own.

The DPA is governed by the laws of the Netherlands, with disputes subject to the Amsterdam District Court (Rechtbank Amsterdam).

The Data Processing Agreement, with Insight and Governance

We provide a Data Processing Agreement under Article 28 GDPR as part of the Insight and Governance agreements. To review it before you sign, write to us.

• Request the DPA or ask a question: dpo@tracehold.com
• Privacy enquiries: privacy@tracehold.com
• General: hello@tracehold.com
• Sub-processors: see the current list
• Post: TraceHold B.V., Werfkade 25, 1033 TA Amsterdam, Netherlands

Related pages: Security, Privacy, Terms, DPA and Subprocessors.