AI Disclosure

Last reviewed: 2026-05-01

TO BRIEF is an AI-native news publisher. This page is the standalone AI-disclosure notice required by Article 50(4) of the EU AI Act (Regulation (EU) 2024/1689) for deployers of generative-AI systems that produce text, images, and other media published to the public. It is intentionally concise so that any reader can understand, at a glance, how this service is produced and who is accountable for it.

What we do

TO BRIEF publishes news briefs — text, images, and translations — that are produced by generative-AI systems. No human editor reviews articles before publication. For full technical detail see /methodology.

Why we do it

AI allows us to cover more of the Greek and European news agenda, with consistent citation discipline, and to translate our work into every official EU language without waiting weeks for human translators. The trade-off is that we lose the safeguards that an experienced newsroom provides. We accept that trade-off explicitly by emphasising post-publication accountability mechanisms instead of prospective gatekeeping.

What the AI actually does

For each brief, AI systems:

• fetch and deduplicate items from curated RSS and Atom feeds;
• cluster related items into stories;
• conduct targeted web research against primary sources, including live web search;
• draft the final headline, body, and inline citations;
• select or generate a headline image, watermarked with Google's SynthID;
• translate the Greek master into every supported EU language for publication on tobrief.eu.

Models used for text synthesis are Claude Opus 4.7 (Anthropic) and Gemini 3.1 Pro (Google), both with web-search grounding. Image generation uses Gemini 3.1 Flash Image with Imagen 3 fallback.

Publisher accountability

The publisher is TO BRIEF Ι.Κ.Ε. (GEMI 188116103000, AFM 803045387), registered at Troon 26, 118 51, Αθήνα, Αττική, Ελλάδα. As a legal entity, the publisher assumes editorial responsibility for all content on the Service. This means:

• we stand behind what is published as the publisher of record;
• we correct errors in a public /corrections log;
• we investigate reader reports submitted to /report under the SLA described in /complaints;
• we publish an annual transparency report covering complaints, corrections, and material model changes.

What to watch for

AI-generated news has characteristic failure modes: fabricated quotes or citations, outdated context despite live search, propagation of source bias, translation drift, and occasional clustering errors that merge unrelated events. Images may misrepresent their subject. Please read /methodology § 5 for the full catalogue of risks.

What you can do

• Verify claims by clicking inline citations — each sentence-level assertion links to the source we relied on.
• Flag issues at /report. Factual errors are prioritised.
• Withdraw reliance for decisions with legal, medical, financial, or safety consequences — seek qualified human advice in those contexts.

Regulatory framing

This disclosure is made under, and is designed to satisfy, the following regimes:

• AI Act (Regulation (EU) 2024/1689) Art. 50(4) — transparency obligation for deployers of generative-AI systems whose output is made public.
• European Media Freedom Act (Regulation (EU) 2024/1083) Art. 6 — ownership, representation, and funding transparency (see /company).
• Digital Services Act (Regulation (EU) 2022/2065) Arts. 16 and 23 — notice-and-action mechanism and annual transparency reporting (see /complaints and /transparency).
• GDPR (Regulation (EU) 2016/679) — data-protection disclosures (see /privacy).

Contact

Questions about AI systems used by the Service, about specific outputs, or about this disclosure: [email protected].