Tech giants respond to European AI rules with enhanced watermarking and labelling, transforming how synthetic content is disclosed and identified across platforms.
Tech companies are moving to make artificial intelligence more visible at the point of use, a shift that reflects both regulation and growing concern about synthetic content crowding out human-made work. According to Nederlandse Dagblad, the timing of Spotify’s new labels and Anthropic’s watermarking is not accidental: the European Union’s AI Act began requiring clearer disclosure from August 2, pushing firms to identify what is human-made and what is generated by AI. Spotify has already said it removed 75 million AI tracks in the past 12 months, a sign of how quickly low-quality synthetic music has accumulated on major streaming services.
Anthropic has now added machine-readable watermarks to text and files produced by Claude, in line with the new European transparency rules. Axios reported that the company is applying the markings globally to models launched in the European Union after August 2, while also acknowledging limits in detection. The exact method is not being disclosed, but researchers say watermarking can be built into the statistical choices a language model makes when selecting words, making the signal hard for ordinary users to notice but detectable by software.
Spotify’s approach is more public-facing. From September, people uploading music will be able to label an artist as an “AI Persona”, and that label will appear on the profile. The company also plans to use SongDNA, a new credits feature, to show whether AI was used in lyrics, composition or other parts of a track. TechRadar reported that Spotify will also flag accounts it believes are likely to be AI personas, although the company says the system is aimed at identity and imagery rather than the music itself.
That distinction matters. MusicRadar reported that Spotify’s previous enforcement has focused heavily on spam, impersonation and deepfakes, because synthetic tracks can exploit the platform’s royalty system with little effort. The company says it wants to improve trust without penalising legitimate artists who use AI as one tool among many. But critics argue that labels based on self-disclosure still leave too much room for ambiguity, especially when a song mixes human performance with machine-written elements.
The broader policy backdrop suggests more disclosure is coming. Axios said 190 organisations had already signed the EU’s transparency code before August, including Google, Meta, Microsoft, OpenAI and Getty Images, and the Dutch report notes that other companies are already using AI labels in retail and publishing. For creators, the issue is not simply whether AI is present, but how much of a work is synthetic and whether audiences can tell. As one Dutch music representative put it, the current labels are a useful start, but the industry will still need a better balance between innovation, authorship and protection for writers and performers.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
