EU AI Act Transparency Rules Are Now in Force: Chatbots Must Disclose, AI Content Must Be Marked

The EU AI Act's Article 50 transparency obligations took effect on August 2, 2026, and they reach far more businesses than the Act's headline high-risk categories. Any AI system that interacts directly with people — customer-service chatbots, AI assistants, conversational interfaces of any kind — must now clearly disclose that it is artificial. Deepfakes and AI-generated text published to inform the public must carry visible markings. Non-compliance carries fines up to €15 million or three percent of worldwide annual turnover, whichever is higher. The timeline includes meaningful relief elsewhere: newly deployed systems must comply immediately, existing systems have a grace period until December 2, 2026, and the high-risk Annex III categories — recruitment tools, credit scoring, education, critical infrastructure — were pushed to December 2027, with product-embedded AI under existing EU safety law moving to 2028. The European Commission adopted interpretive guidelines on July 20, giving companies a concrete reference for what adequate disclosure looks like. The compliance work for most businesses is unglamorous inventory-taking: enumerate every surface where AI talks to a human or where AI-generated content is published, and check each for disclosure. The subtler exposure sits with companies that quietly adopted AI content generation — product descriptions, news-style articles, marketing copy presented as editorial. Under Article 50, publicly informative AI text needs marking, and the reputational cost of being caught unmarked may exceed the regulatory one: platforms are already learning that audiences forgive disclosed AI content far more readily than discovered AI content.