Skip to content
NewsRegulation

EU AI Act Article 50 transparency rules take effect, requiring chatbots to identify as AI and deepfakes to carry labels

· by Pondero Newsdesk

The short version

Article 50 of the EU AI Act became enforceable on August 2, 2026, obligating providers and deployers serving EU users to disclose automated interactions, label synthetic content, and mark deepfakes. Fines reach 15 million euros or 3% of global annual turnover.

EU AI Act Article 50 transparency rules take effect, requiring chatbots to identify as AI and deepfakes to carry labels

Article 50 of the EU AI Act became enforceable on August 2, 2026. Every chatbot serving EU users must now tell the person it is talking with that the system is AI. Deepfakes need machine-readable metadata. News articles generated by AI without human editorial review need labels. Teams that missed the date are already exposed to fines.

What

Per the European Commission's transparency fact page, the obligations split between two actors: providers (companies that build AI systems) and deployers (companies that embed those systems in products and services).

Providers face two requirements. First, they must design chatbots and other automated interaction systems so that users know they are dealing with AI, not a person. Second, machine-readable marks are required on any text, image, video, or audio that a generative AI system produces or manipulates, unless the output is purely assistive editing that does not substantially alter the input.

Deployers carry separate disclosure duties. Any service running an emotion-recognition or biometric-categorization system must tell the people being analyzed. Platforms that publish deepfakes, defined as AI-generated or manipulated content depicting real persons, places, or objects in a way that could be mistaken for authentic footage, must display a clear and perceivable label. That same labeling requirement covers AI-written text on matters of public interest that has not gone through human editorial review.

The rules apply to any company serving EU users, regardless of where the company is incorporated. OpenAI, Anthropic, Google, Meta, and Microsoft are all in scope.

Fines run up to 15 million euros or 3% of worldwide annual turnover, whichever is higher, with proportionality provisions for smaller companies. One grace period remains: generative AI systems already on the market before August 2, 2026 have until December 2026 to meet the machine-readable marking requirement, per the AI Omnibus amendment to Article 50(2).

Why it matters

The chatbot disclosure obligation cannot be offloaded to a model vendor. The duty falls on whoever presents the AI interface to the end user, per TechTimes. That makes every SaaS product, support widget, and customer-service flow that touches EU customers an active compliance decision right now, not a backlog item.

The deepfake labeling requirement reaches beyond synthetic media companies. Marketing, content, and creative tools that use generative AI to produce or modify imagery and video depicting real people need to assess whether their outputs fall under the Commission's definition, which covers both created and manipulated content.

On the same day the rules took effect, the EU called for closer monitoring of high-risk AI systems following the July 2026 containment incidents involving OpenAI and Anthropic, per Global Banking and Finance. Regulators are assessing whether those incidents carry separate obligations under the Act's enforcement framework, distinct from the Article 50 transparency chapter.

What to watch next

The first formal warning or investigation notice from a national market surveillance authority or the AI Office will set the enforcement tone for the rest of 2026. Commission guidelines classifying high-risk AI applications are due later this year and will define the obligations wave that follows Article 50.

Sources