EU AI Act enforcement starts August 2 with new transparency rules for chatbots and deepfakes
EU AI Act enforcement begins August 2, adding transparency duties for chatbots, deepfakes and AI-generated content.
The European Commission says the next major phase of the EU AI Act begins on August 2, 2026, when its AI Office and national authorities start enforcing the law and new transparency obligations take effect. The move turns a two-year-old regulatory framework into a live compliance regime for companies offering or deploying AI systems in the European Union.
The immediate rules focus on disclosure. Certain interactive AI systems, including chatbots, must make clear to users that they are dealing with AI rather than a person. AI-generated or AI-altered images, audio and video, including deepfakes, must be labelled, and generated or altered content must carry machine-readable marks intended to make automated detection easier.
What changes now
The Commission framed the transparency measures as a way to reduce deception and manipulation while giving businesses clearer obligations. It also pointed to a transparency code of practice and said more than 180 organisations have signed up to the first published list of participants. That code is designed to help operationalise rules for AI-generated content, although the legal obligations come from the AI Act itself.
Enforcement will be shared between the Commission's AI Office and authorities in EU member states. The Commission's general AI Act guidance says the AI Office has powers over general-purpose AI models, including the ability to request technical documentation, evaluate models, require corrective measures and issue fines for non-compliance.
Why it matters
The August 2 start date is significant because it shifts the AI Act from preparation into supervision for a broader set of providers and deployers. Companies that ship chat interfaces, synthetic media tools, AI assistants or systems that publish public-interest text will need to review how users are notified, how generated media is labelled and whether those notices survive through normal distribution channels.
Some obligations still arrive later. The Commission's timeline says high-risk AI systems in sensitive areas such as biometrics, critical infrastructure, education, employment and migration are scheduled to face strict obligations from December 2, 2027. Systems embedded in regulated products such as lifts or toys have an extended transition period until August 2, 2028.
For technology vendors outside Europe, the practical impact is still global. Services available in the EU may need product changes, documentation and governance processes that are difficult to isolate by region. As with privacy regulation before it, the EU's AI rulebook could shape defaults for companies that prefer one compliance model across multiple markets.
Sources
Cover photo by Sora Shimazaki on Pexels, used under the Pexels License.
Comments (0)
Leave a Comment