Picture a piece of text that, read by a human, looks perfectly natural, but carries, hidden among the words, a statistical signature only specific software can detect. This isn’t science fiction. Since 2 August 2026, it’s what the European Union legally requires of anyone generating content with artificial intelligence within its borders.
The obligation comes from Article 50 of the EU AI Act, which took full effect on that date, and it’s the first continent-wide attempt to systematically track what an AI produces, whether text, image or audio. The guiding principle, according to the Code of Practice that accompanies the rule, is that no single marking technique is enough on its own. Providers must combine digitally signed metadata, imperceptible watermarking and, as a fallback, fingerprinting or logging systems.
To see how this works in practice, take text. A technology like Google’s SynthID works by nudging word choices along a pseudorandom pattern: the text reads entirely normally, but across enough words a statistical signature builds up that verification software can recognise. For images and audio, the principle is similar, though the signal is embedded invisibly or inaudibly directly into the file.
Anthropic has said that Claude models launched on or after 2 August 2026 support marking from launch, embedding invisible watermarks into text the chatbot generates. It’s a sign that the sector’s major players are moving in step with the regulatory deadline rather than scrambling to catch up afterward.
The penalties for non-compliance aren’t symbolic. Fines can reach €15 million or 3% of a company’s global annual turnover, whichever is higher, a figure designed to bite whether the offender is a startup or a tech giant. It marks a shift from the lighter-touch approach the EU had taken, until recently, toward tracking AI-generated content.
Research for this article turned up no direct, verifiable link between this regulation and any Canary Islands company or institution. This is, in every sense, a global regulatory shift that touches anyone using generative AI tools, from a tourist writing a caption for a social post to a newsroom leaning on AI for a first draft.
