As the European Union’s flagship artificial intelligence law moves from theory to implementation, AI labeling of content that appears to be authentic becomes mandatory for companies that build or use artificial intelligence. From August 2, providers and users across the Union will face binding transparency rules covering deepfakes, synthetic media and machine-generated text, a change that platforms have been preparing for since YouTube began automatically labeling AI videos.
See also: Europe Banks: Mandatory refunds to phishing victims?
The basic idea is simple: If a piece of content can pass as real, people have the right to know that it was machine-generated, and the onus now falls on the companies that created it.
The rules divide the project into two parts. Providers, the companies that build genetic systems, must embed machine-readable tags so that images, audio, video, and text can be detected as being created or modified by AI. Users take on the visible part. Anyone who uses AI to create a deepfake or publish machine-written text on matters of public interest must disclose that origin to the public in a clear and timely manner.
Intent does not absolve you of the obligation. The obligation to label a realistic deepfake applies regardless of whether the creator intended to deceive, closing a loophole that the “ we were just kidding ” excuse could open . Chatbots are also covered.
When a person begins interacting with an AI system, they should be informed at the moment of contact, not in the fine print of a terms of use page that no one reads. There are real exceptions. Content that is clearly unrealistic or fictional does not fall under the rules, and AI-generated text escapes flagging if a human with genuine editorial responsibility has substantively reviewed it.
Creative work is treated more lightly. Artistic and satirical content must still be revealing, but only in a way that doesn’t ruin the experience of the work, a credit to the filmmakers and artists. Penalties give the rules “teeth.” Violations can result in fines of up to 15 million euros or 3% of a company’s global annual turnover, whichever is greater, enough to draw attention to Silicon Valley and Brussels.
See also: Microsoft: Mandatory use of MFA for access to admin portals
The timing is not yet set. The obligations officially go into effect on August 2, but a separate simplification package has proposed pushing back the deadline for machine-generated labeling until December, part of a broader effort to water down parts of the AI law. The industry is pushing for limits. Advertisers, among others, have argued that AI-generated ads should be exempt from the transparency rules, a lobbying battle that has been waged alongside the law’s implementation.
The Commission has tried to add clarity. It published draft guidelines in May and is finalising a code of practice on how to mark and label AI content, which companies can sign up to show they are compliant. The reach goes beyond Europe.
Because global platforms rarely build separate systems for a single market, the EU’s labeling regime is likely to shape how AI content is labeled globally, much as privacy law previously did. Enforcement will be the hard part. Watermarks can be removed, metadata can be lost, and machine-written text is notoriously difficult to detect, so the rules will only be as strong as the tools that enforce them.
Deepfakes are the most intense test. The EU has moved separately to ban non-consensual personal deepfakes, and the labeling regime is intended to catch the wider universe of synthetic media before it completely erodes trust. The stakes go beyond compliance. Europe is betting that mandatory transparency can preserve some common sense of what is real online, at a time when genetic tools make it easy to produce convincing fakes.
See also: Windows 11: Mandatory SMB encryption for outgoing connections
Whether it will work is an open question. The law is the most ambitious attempt yet to flag the synthetic world, but its success depends on whether detection can outpace creation, a race that counterfeiters have so far been winning.
