Version: 2 October 2026.
Regulation (EU) 2024/1689 (the AI Act) applies to AI systems. PassportAI is an AI system. In this statement we explain how we classify PassportAI under the Regulation and which safeguards are built into its design.
PassportAI is intended for professionals at and for financial institutions — including compliance officers, in-house lawyers, risk managers and their external advisers — as a tool for regulatory research: it searches EU legislation and guidance and the national financial regulation, supervisory guidance and case law of the countries covered, analyses comparable provisions and decisions and, on that basis, generates a research result with references to its sources (for example CELEX and ECLI numbers, article references and guideline references). The legal assessment of that result and any decision based on it remain the work of the user.
PassportAI is not intended (i) to apply the law to a specific case independently or to determine the outcome of (judicial) decision-making, (ii) to take or prepare automated decisions about persons, or (iii) to be used without human oversight and verification. Use outside this intended purpose is in breach of the terms and conditions.
Annex III, point 8, of the AI Act (administration of justice and democratic processes) classifies as high-risk: AI systems intended to be used by or on behalf of judicial and arbitral authorities to assist in researching and interpreting facts and the law and in applying the law to a concrete set of facts. For use by compliance officers, lawyers and other non-judicial professionals, PassportAI falls outside this description.
Should PassportAI nevertheless be used by or on behalf of judicial or arbitral authorities — to look up and analyse comparable decisions in preparation for their own judgement — it touches on the scope of point 8(a). For that use, the exception in Article 6(3) of the Regulation applies: PassportAI performs a preparatory task for an assessment that the human carries out (point (d)) and improves the result of a human research activity (point (b)), without materially influencing the outcome of decision-making — every answer states the sources on which it rests, verification of those sources by the user is a condition of use and the decision itself remains entirely human work. PassportAI does not perform profiling of natural persons.
In our assessment, PassportAI is therefore not a high-risk AI system within the meaning of the Regulation. We keep this assessment documented and review it whenever the service or its intended purpose changes materially.
Users know that they are working with an AI system: answers are expressly presented as AI-generated, with references to their sources and with the standing warning that PassportAI does not provide legal advice (Article 50 of the AI Act).
PassportAI uses a general-purpose AI model with openly available weights (open-weight) from a third-party provider, accessed via European inference infrastructure with zero retention (Seeweb S.r.l. / Regolo.ai, with a European fallback — see the privacy policy); there is no model provider outside the EU in the chain. Converting questions and source fragments into search vectors is done with an open-source model (bge-m3) via OVHcloud AI Endpoints (OVH SAS, France), likewise without storage. The obligations that the Regulation imposes on the provider of such a model rest with that provider; we take the information supplied by the model provider into account in our own assessment.
Questions about this statement can be asked via the Support button at the top of this page.