During the Copyright Royalty Board’s (CRB) comment period over the negotiated settlement establishing the U.S. mechanical rate for songs in the 2028-2032 five-year term — including applying a cost-of-living adjustment (COLA) each year to the current rate level — a formidable group of companies and some well-known music trade organizations (which include those who negotiated the settlement) came forward in support.
However, comments to the CRB judges from a number of organizations whose websites or public statements claim they represent a large number of songwriters and publishers, as well as the usual group of respected U.S. music industry contrarians, are equally vocal in saying they do not support the settlement. Many of them urge the CRB judges to establish a higher rate than what the settlement is proposing.

– Ed Christman, Billboard

AI companies are now required to make model outputs detectable as AI-generated under Article 50 of the EU AI Act – or, in other words, watermarked.
The major AI companies have all signed the additional voluntary Code of Practice. OpenAI has still not announced what measures it will be using, despite having built and not released a text-watermarking capability years ago. Google already watermarked AI generated images and videos, which it has since extended to text. Anthropic, however, has triggered the biggest headlines, unveiling text watermarks across all models (including older ones) and signed C2PA metadata in generated files, worldwide – not just in the EU.

– Hanna Kahlert, MIDiA

There have been a number of high-profile legal rulings and settlements around AI training in the past year, and for non-lawyers, figuring out what precedents they set and what impacts they might have can be a slog.
Thanks are due, then, to Berklee College of Music Professor Pete Dyson, who has published ‘The AI Training Wars’ – a detailed but clear Substack post breaking down some of those key rulings and settlements, and digesting the implications.
It explores three of the most important themes that have emerged from cases involving Anthropic, Suno, Meta, and Google, and explains why there is still a lot of uncertainty about the legals around training AI models on copyrighted material.

– Stuart Dredge, Music Ally

Suno is introducing deliberate friction to pump the brakes on the AI-powered superabundance it helped create. By implementing download caps (per Variety), the company is putting guardrails around how much music users can output.
Streaming services have removed millions of suspected ‘spam’ tracks, and TechRepublic reports Spotify is introducing AI persona labelling. With Suno also signing a licensing deal with BMG (according to Music Business Worldwide), all signs point to a world where AI is seen less as a bull in a china shop and more as an opportunity that can viably be integrated, governed, and monetised by the music industry. Ultimately, as the integration of AI unfolds, the industry will increasingly need to solve the cultural problems created by superabundance, alongside reaping the functional benefits AI can deliver.

– Kriss Thakrar, MIDiA

Apple Music sent an email, obtained by Billboard, to music industry partners on Thursday (Aug. 20), announcing its Made With AI labeling initiative. Later this year, the platform notes that any song that is tagged as being “materially generated using AI” by content providers, like record labels and distributors, will be visibly labeled on the service for users to see.
While it is unclear exactly when these labels will be launched, Apple Music notes that they will appear on the service “later this year.” The news builds on the company’s March announcement of AI Transparency Tags — a system requiring content providers to disclose when and where AI is used in a “material” way, like for the creation of artwork, track, composition and music video.

– Kristin Robinson, Billboard