{"id":656,"date":"2026-07-22T15:01:24","date_gmt":"2026-07-22T06:01:24","guid":{"rendered":"https:\/\/onepress.co.kr\/index.php\/briefing\/2026-07-22-eu-ai-act-transparency-article-50-en\/"},"modified":"2026-07-22T15:29:40","modified_gmt":"2026-07-22T06:29:40","slug":"2026-07-22-eu-ai-act-transparency-article-50-en","status":"publish","type":"briefing","link":"https:\/\/onepress.co.kr\/index.php\/briefing\/2026-07-22-eu-ai-act-transparency-article-50-en\/","title":{"rendered":"EU AI transparency rules start applying on 2 August"},"content":{"rendered":"

2026-07-22 14:20 KST<\/strong><\/p>\n

The European Commission published Article 50 transparency guidelines on 20 July 2026, and the obligations start applying on 2 August 2026. The core change is clearer notice for direct AI interaction, marking of synthetic content, and disclosure duties for deepfakes and certain public-interest AI text.<\/p>\n

What changes<\/h2>\n

The European Commission published guidelines for Article 50 of the AI Act on 20 July 2026. The transparency obligations start applying on 2 August 2026 and focus on whether people are directly interacting with AI, whether AI-generated or manipulated content must be marked, and when deployers must disclose deepfakes or certain AI-generated text on public-interest matters.<\/p>\n

Who is covered and when<\/h2>\n

The rules cover several distinct situations. AI systems that directly interact with people must tell them they are interacting with AI unless that is already obvious. Providers of in-scope systems generating synthetic audio, image, video or text must make outputs machine-readable and detectable as artificially generated or manipulated. Deployers can also face disclosure duties for deepfakes, emotion recognition, biometric categorisation and some AI-generated public-interest text.<\/p>\n

\"Generated
Generated image: ONEPRESS created this image to explain the notice and marking structure of EU AI transparency rules. It is not a screen from a specific service.<\/figcaption><\/figure>\n

Practical impact<\/h2>\n

The practical effect may reach beyond the EU. Companies offering AI features in the EU market or to EU users need to review chatbot entry points, metadata and marking for generated media, and human-review procedures for informative text. For users, the rule is meant to make three questions easier: am I interacting with AI, was this media made or altered by AI, and has this public-interest text been reviewed by a responsible person?<\/p>\n

What is still unsettled<\/h2>\n

The rules do not mean every AI output receives the same label. The Commission Q&A lists outputs that can fall outside the marking obligation, including short sequences of numbers or symbols, source code, machine-to-machine outputs not exposed to people, and some closed-loop industrial or production environments. The boundary between standard editing assistance and generation or manipulation also depends on context.<\/p>\n

Next dates and primary sources<\/h2>\n

The next checkpoint is enforcement practice after 2 August 2026 by national market surveillance authorities and the AI Office. The Code of Practice is voluntary, but Article 50 obligations are legal obligations. Providers and deployers should compare the guidelines, Q&A and Code of Practice when documenting notices, marking methods, human review and any exemption they rely on.<\/p>\n

European Commission Article 50 guidelines and application date<\/a><\/p>\n

Commission Q&amp;A on scope and exemptions<\/a><\/p>\n

Commission release summarizing the obligations<\/a><\/p>\n

Transparency Code of Practice and its voluntary status<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

The European Commission published Article 50 transparency guidelines on 20 July 2026, and the obligations start applying on 2 August 2026. The core change is clearer notice for direct AI interaction, marking of synthetic content, and disclosure duties for deepfakes and certain public-interest AI text.<\/p>\n","protected":false},"featured_media":0,"template":"","meta":[],"class_list":["post-656","briefing","type-briefing","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/onepress.co.kr\/index.php\/wp-json\/wp\/v2\/briefing\/656","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/onepress.co.kr\/index.php\/wp-json\/wp\/v2\/briefing"}],"about":[{"href":"https:\/\/onepress.co.kr\/index.php\/wp-json\/wp\/v2\/types\/briefing"}],"wp:attachment":[{"href":"https:\/\/onepress.co.kr\/index.php\/wp-json\/wp\/v2\/media?parent=656"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}