EU AI Act: Transparency duties and GPAI rules in force since 2 August
Chatbots must reveal they are AI, deepfakes and AI content must be labeled – and providers of large AI models now fall under the direct oversight of the European Commission.
Illustration · AI-generated (AI IN LIFE)
At a glance
- In force since: 2 August 2026
- Oversight: EU AI Office + national authorities
- Duties: chatbot disclosure, deepfake/AI-content labeling, GPAI transparency
- Over 180 organizations signed the voluntary Code of Practice
Since 2 August 2026, the European Commission has been enforcing central parts of the AI Act. For businesses and users in Europe it is the most tangible step of AI regulation so far: anyone interacting with an AI system must now be able to recognize it.
Three things apply in concrete terms. First, transparency: interactive AI systems such as chatbots must disclose that you are talking to a machine, not a human. Second, labeling of synthetic media: deepfakes and AI-generated or -altered content must be marked as such – ideally with machine-readable signatures so platforms can detect them automatically. Third, oversight of General-Purpose AI (GPAI): providers of large foundation models face direct obligations for the first time, supervised by the newly created AI Office together with national authorities.
All of this is flanked by a voluntary GPAI Code of Practice. According to the Commission, more than 180 organizations have already signed the code of practice on transparency of AI-generated content – a signal that many providers would rather commit early than wait for sanctions.
For newsrooms like AI IN LIFE this is no rupture but a confirmation: we already label our shows and reports visibly as AI productions and link the sources we use. What the EU now makes mandatory is our standard.
What does it mean in practice? Companies using AI in customer contact – from WhatsApp automation to voice agents – should check now: does the user know they are talking to an AI? Are generated images, voices and videos labeled? Those who set this up cleanly are not just compliant, they earn trust.
FAQ
Will every chatbot have to say it is an AI?
Yes. Interactive AI systems must inform users they are communicating with a machine and not a human, unless it is obvious.
Do AI-generated images and videos have to be labeled?
Yes. AI-generated or -altered content and deepfakes must be marked as such, ideally with machine-readable signatures.
What is GPAI?
General-Purpose AI refers to broadly usable foundation models such as large language or image models. Their providers face direct EU obligations for the first time since 2 August.
