AI transparency: what Article 50 requires from companies on August 2
Labeling deepfakes, disclosing chatbots, machine-readable marking: a full look at the obligations now taking effect in Europe.
On August 2, 2026, a new layer of the European AI regulation becomes binding. It is Article 50 of the AI Act, devoted to transparency. In practice, several kinds of AI-produced content will have to be clearly identified as such, and the companies deploying these systems now face precise obligations.
What Article 50 says
The text targets several situations. Conversational systems, such as chatbots, must warn users that they are talking to a machine, not a human. Synthetic content (images, audio, video, and also text published on matters of public interest) must carry a marking indicating its artificial origin. Providers of generative models must additionally embed a machine-readable mark so the content can be detected automatically, except when the AI only performs a minor assistive role that does not really alter the content provided.
One principle runs through it all: the information must be given at the moment a person encounters the content, not buried in terms of service or a secondary screen.
Deterrent penalties
Compliance is not optional. Breaches of the AI Act can be penalized, in the most serious cases, with fines of up to 15 million euros or 3% of a company’s worldwide turnover. To support compliance, the European Commission has published a voluntary code of practice detailing concrete marking and labeling methods.
Why it is a turning point
Until now, flagging AI-made content was mostly a matter of goodwill. That is over: transparency becomes a legal requirement. For platforms, media outlets and companies that use chatbots or publish generated content, this means adapting interfaces, adding disclosures and, for providers, embedding technical marking.
For the public, the change will show over the coming weeks: more and more notices like AI-generated content, or you are talking to an automated assistant, will appear. It is a useful cue to know who, or what, you are dealing with.
At IAscope, some content is written with AI assistance then reviewed and approved by a human, so we apply this same transparency logic to ourselves.