What actually became law on August 2?
Article 50 of the EU AI Act (Regulation (EU) 2024/1689) stopped being a future deadline and became enforceable across the European Union on August 2, 2026. The rule bundles two separate obligations that a lot of coverage blurred together. Anyone deploying AI that produces a deepfake, meaning audio, video or image content realistic enough to make a listener believe something happened that did not, has to disclose that the content is artificially generated. That duty applied on schedule. Providers of the underlying AI tools also have to embed machine-readable marking so their output is detectable as AI-made, but that piece got a four-month transitional window, pushed to December 2, 2026, for systems already on the market before the deadline. Miss either obligation and national market-surveillance authorities can levy fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher.
None of that targets a producer directly. It targets companies: the AI tool maker, and whoever deploys the content commercially, which in music means the distributor and the platform. That is exactly why the practical weight of this law is landing on the businesses that sit between a track and a listener, not on whoever made the track.
What does this mean for a Beatport or SoundCloud upload?
The mechanism carrying a disclosure from a bedroom studio to a listener's app is boring and specific. DDEX, the metadata standard the industry already uses to move release data between labels, distributors and streaming services, now carries fields for AI involvement, covering vocals, instrumentation, composition, post-production and lyrics. Spotify adopted the standard in 2025. Fill the field in accurately at upload and it travels through to song credits and platform labels automatically. Leave it blank and, by design, no tag travels downstream at all: there is no automatic fallback catching it on the platform end.
Beatport moved first among the dance-specific platforms. Ten days after Article 50 became enforceable, it expanded its partnership with fraud-detection company Beatdapp and rewrote its Content Guidelines: fully or majority AI-generated tracks are now withheld at ingestion, with rightsholders notified directly, while AI-assisted tracks that stay majority human-made get tagged and passed on to human curators.
"Beatport exists to serve DJs and empower artists and labels. There is a difference between a tool that assists human creation and a system that replaces it entirely. Beatport is built on the former." - Beatport CEO Matt Gralen
The policy followed a Beatport user survey in which 60% of respondents said they would not play AI-generated tracks, 77% of those citing a preference for supporting human creators. SoundCloud, by contrast, has not published a comparable AI-disclosure policy. An undisclosed AI track can still reach SoundCloud listeners with no flag at all, at least for now, which means the same file Beatport withholds at its door can upload clean somewhere else.
Who actually eats the risk if the metadata is wrong?
Here is the split that matters for anyone running a label or submitting a track: the EU's fines are aimed at the distributor and the platform, the businesses now legally on the hook for the metadata's accuracy. Chartlex, the trade outlet that first flagged this deadline, put the producer's side of it plainly: the realistic exposure for an independent artist is takedown and demonetization rather than a regulator's fine.
But that split runs downhill. A distributor staring down 3% of global turnover in potential exposure has no reason to give an ambiguous upload the benefit of the doubt. The AI-disclosure field on a Beatport or DistroKid upload form is no longer a courtesy checkbox, it is the paperwork a company fills out to protect itself from a European regulator, and a producer who leaves it blank, or fills it in wrong, is the one whose track gets pulled to cover that exposure.
