EU AI Act — Article 50 Transparency
Article 50 of the EU AI Act has applied since 2 August 2026. This page explains, in specifics, how Tribe AccredAI discloses AI interaction, labels AI-assisted content, marks exported documents, and records human approval.
Scope and our role
Article 50 of the EU AI Act (Regulation (EU) 2024/1689) has applied since 2 August 2026. It reaches any organization whose AI outputs are used inside the EU, regardless of where the organization is established. Tribe AccredAI is a deployer of third-party general-purpose models accessed through a managed AI gateway — we do not develop or place a generative model on the market. That distinction shapes which duties fall to us and which fall to the institution using the platform.
Article 50(1) — You always know when you are talking to AI
Ada, our accreditation assistant, carries a persistent disclosure in its header before the first message is sent, and the first response in each session repeats that the answer was AI-generated and should be verified. The same disclosure appears on our public FAQ assistant, which is reachable without signing in.
Article 50(2) — AI-assisted content is labelled and marked
AI-drafted narratives, AI-suggested evidence mappings, and AI-derived summaries carry a visible label inside the product. The label distinguishes unreviewed AI drafts from AI-assisted content that a named human has approved. Documents exported from the platform embed machine-readable document properties identifying AI-assisted content, so the marking survives outside our interface.
Article 50(4) — Human editorial responsibility
Accreditation self-studies are frequently published or filed with public bodies. Our workflow requires an authorised institutional user to review and approve AI-assisted content before finalization, and the approval — who and when — is recorded. Exported documents state plainly that portions were drafted with AI assistance and reviewed by an authorised institutional user.
Article 50(5) — Clear, distinguishable, accessible disclosure
Disclosures are presented at the first interaction, in plain language, in the same interface as the content they describe, and are readable by assistive technology. They are not buried in terms of service.
Auditable record for your own compliance file
Workspace administrators can open the AI activity record inside the platform: a metadata-only log of which feature used AI, for what purpose, which model family, who triggered it, and who approved the result. It stores no prompts, no evidence content, and no student records. Institutions can export it as a supporting table for procurement, internal audit, or a regulator request.
Accrediting organizations operating internationally
Tribe AccredAI also serves accrediting organizations whose review activities cross borders — including accreditors serving international councils and institutions in the EU. Accreditor-side review, clarification, and decision workflows are human-driven and currently use no generative AI. When an institution sends a formal submission, the metadata-only AI provenance (which sections were AI-assisted, the model family, and the approving human) travels with the locked submission snapshot, so reviewers anywhere can see and rely on the same AI disclosure the institution recorded. Any future AI assistance on the accreditor side will carry the same labelling and audit record described on this page.
Data used for AI, and model training
Institutional content is processed only to deliver platform functionality. We do not use customer content to train public AI models. AI processing of identifiable student records is off unless an administrator explicitly enables it. Model providers used for AI processing are named on our Subprocessors page.
What remains the institution's responsibility
Your institution is a deployer in its own right. Where you publish AI-assisted text to inform the public, you must keep the human review documented, retain editorial responsibility, and disclose AI assistance where the exemption does not apply. Tribe AccredAI supplies the labelling, provenance, and audit record to make that straightforward — it does not transfer the obligation.
| Obligation | Where it applies in the platform | How we meet it |
|---|---|---|
| 50(1) Interactive AI | Ada assistant, FAQ assistant | Persistent header disclosure plus a first-reply notice in every session |
| 50(2) Synthetic content | AI-drafted narratives, AI-suggested evidence mappings | Visible AI label in-product; machine-readable marking in PDF/DOCX properties |
| 50(4) Published text | Self-study reports, evidence appendices, submission manifests | Recorded human approval, provenance section, and a disclosure line on the document |
| 50(5) Presentation | All of the above | Plain language, first interaction, accessible markup, not buried in terms |
| Evidence of oversight | Workspace administration | Metadata-only AI activity record, exportable for audit |
This page describes product behaviour. It is not legal advice and is not a certification of conformity. For AI governance or procurement questions, contact legal@accredai.com.