Privacy notice.

In one paragraph.

Trustgent collects the minimum personal data needed to operate the index. Our production servers are currently hosted in the United States (detail under "Where data is processed"). We never sell personal data. We log access for security and abuse detection only. EU GDPR and UK GDPR apply; we honour subject access, rectification, erasure, and portability requests within 30 days.

What we collect.

• Provider profiles from public sources:names and business roles of people connected to a listed provider (founders, named contacts), collected from the provider's own site, published case studies, and public registries. See "Profiles built from public sources" below.
• Newsletter subscribers:email address + consent record. Stored on our production database (see "Where data is processed").
• Registered accounts: email address, role (buyer / provider / admin), basic profile.
• Verified customer raters:email address (domain-matched, encrypted at rest) and the rating submitted. A published rating shows the stars, the outcome text, and the rating organisation's domain; the rater personally is anonymised.
• Research questionnaires (optional):for buyers: role, company size, industry, AI maturity stage, timeline, and top selection criterion. For providers (the builder pulse): revenue model, AI revenue share, delivery constraint, model stack, pipeline outlook, and acquisition channel. All in coarse bands; every field can be skipped. Used only as described under "Published research" below.
• Provider submissions: identification of the individual claiming the listing, plus the submission contents the provider chooses to publish.
• Server logs: IP address, user agent, request path, timestamp. Retained 30 days; aggregated, non-personal metrics retained indefinitely. Used for security, abuse detection, and traffic understanding only.

Profiles built from public sources.

A provider profile can exist before the provider has contacted us; it is then marked "not yet verified — claim to verify". That label means we have not yet verified evidence for this provider. It is not a judgment about quality. Profiles are built only from genuinely public professional sources — the provider's own website, published case studies, public business registries — and we record where every data point came from and when we collected it. We never collect from behind logins, paywalls, or restricted-visibility settings. Where a profile identifies a person (a founder, a named contact), our legal basis is legitimate interest (Art 6(1)(f) GDPR): operating a public reference that helps buyers assess providers. You can object at any time via the profile's correction link or privacy@trustgent.com; objections receive a documented, individual decision, and an objection to receiving email from us is always honoured without any balancing, same day, permanently.

Notification emails.

We may send one email to a provider's published business address to say its profile is live and can be claimed. That email is a commercial message and also serves as the notice that we have collected the data described above: it states the source, links this notice, and carries a one-click unsubscribe. Opting out is honoured the same day and recorded permanently (as a hash, so it survives even an erasure request). We only send such email in countries whose law permits it, under a country-by-country rule set, and we keep a record of every send.

Published research and aggregates.

Trustgent publishes research built on the verified index, for example the State of AI reports. Two kinds of numbers appear in it, with different bases:

• Observed counts from the public verified index: how many providers hold which verification level, in which category and country. These are counts of public listings, not personal data. Legal basis: legitimate interest.
• Self-reported answers from the optional research questionnaires (the buyer questionnaire and the provider builder pulse). Submitting an answer is your consent to this use; every field can be skipped, and you can withdraw consent at any time via privacy@trustgent.com. Legal basis: consent. Answering never affects any provider's verification level or rank.

Self-reported answers are published only as aggregates. Any group smaller than three respondents is folded into an "other" bucket, so no individual answer is readable from a published number. Answers are never published individually, never linked to a name, a company, or a provider listing, and never affect any provider's rank or verification level. Raw answers live on your lead record and are deleted with it on an erasure request; already-published aggregates remain, since they contain no personal data.

What is public, what stays private.

Provider listings and approved verification evidence (the claim, the source link, the excerpt) are public: that is the product. Everything else stays private: evidence-vault artifacts, reviewer notes, buyer briefs and quotes, deal records, and rater email addresses (encrypted at rest). Customer ratings are published after review as stars, outcome text, and the rating organisation's domain, with the rater personally anonymised. Ratings are moderated against our published attestation code of conduct, which applies identically to positive and negative ratings.

Where data is processed.

Trustgent is operated by a Dutch company (Rustenhoven Management B.V., Amsterdam). Production infrastructure — the application and its database — runs with Hetzner Online GmbH, a German hosting provider, at its United States datacenter in Hillsboro, Oregon. Processing is governed by our data-processing agreement with Hetzner, under which third-country processing must satisfy Articles 44–46 GDPR and Hetzner must evidence the implemented safeguards on request. We additionally maintain a documented transfer impact assessment for this setup, with supplementary technical measures (TLS everywhere, field-level encryption for sensitive data, access restricted to named accounts). A summary of the assessment is available via privacy@trustgent.com. An earlier version of this notice said data was stored in the EU — that was incorrect, and we corrected it the day we established the facts. Your rights (access, erasure, portability, objection) do not depend on where the servers stand.

Legal bases.

Performance of a contract (account, plan, verification flows), consent (newsletter, the optional research questionnaire, marketing cookies if any), legitimate interest (profiles from public sources, provider notification emails, security logs, abuse detection, aggregate statistics from the public verified index), legal obligation (tax records, regulatory reporting).

Sub-processors.

Hosting: Hetzner Online GmbH (Germany), whose approved subcontractors for the US location are Hetzner US LLC and the colocation operators NTT Global Data Centers Americas, Inc. and QTS Investment Properties Hillsboro, LLC (per Appendix 3 of Hetzner's data-processing agreement). Beyond hosting we use the minimum sub-processor set required to run the site; a current list is maintained at privacy@trustgent.com on request and updated within 30 days of any change.

Your rights.

Subject access, rectification, erasure, portability, restriction, objection. Email privacy@trustgent.com with the request. We respond within 30 days. You may also complain to your local data-protection authority; for the Netherlands that is the Autoriteit Persoonsgegevens.

Cookies.

The site uses Plausible for analytics, which is cookie-free and privacy-preserving. We do not use third-party advertising trackers. Strictly-necessary cookies for authenticated sessions will be documented here as they are introduced.

This policy will be revised as the platform adds features. Material changes are emailed to account holders and announced on the homepage. Version history: v2.2, 17 July 2026 (profiles from public sources, notification emails, transfer safeguards named, sub-processors named); v2.1, 11 July 2026 (data-location correction: production is US-hosted, not EU as previously stated); v2, 10 July 2026 (published research, public/private boundary, rater description corrected); v1, 27 June 2026.