According to Music Business Worldwide, the US Department of Justice has backed OpenAI’s fair use position in a copyright case about training on texts, a notable government intervention that does not directly decide music disputes. Separately, Music Business Worldwide reports that Canada’s SOCAN has sued Suno over alleged copying of 150 songs. This week therefore offers legal direction, not legal clarity, alongside quieter progress in tools and rights administration.
Two unsettled signals for training law
According to Music Business Worldwide, the DOJ argues that training on copyrighted works without a licence can be fair use, although its filing concerns text rather than music and is not a judgment in the music cases. The intervention gives AI developers a federal argument, but music companies should treat it as a potentially influential interpretation rather than a direct answer to their disputes with music generators.
Meanwhile, Music Business Worldwide reports that SOCAN alleges Suno copied 150 songs from its repertoire and is seeking damages and injunctions. That claim remains unproven, yet the case is useful because a collecting society is framing training as a repertoire wide enforcement issue on behalf of members. The two legal moves point in different directions, leaving licensing strategy and evidence as central questions rather than settled doctrine.
AI tools move closer to real workflows
According to The Verge, Roland’s Melody Flip generates simple loops from choices including genre, note density, tempo and key, with less control than full song platforms. That restraint is the interesting product decision. Roland is presenting generative AI as a source of musical material inside a making process, rather than as a substitute for the entire process from prompt to polished track.
In another workflow category, AudioShake says AudioShake’s Multi Speaker 2.0 separates overlapping voices into individual tracks and improves transcription accuracy when people speak simultaneously. Its importance for AI music is indirect but practical. Better dialogue separation can make interviews, documentaries and spoken archives easier to edit, search and repurpose, showing that valuable audio AI is often infrastructure rather than generation.
Royalty infrastructure has real financial scale
According to CISAC, Brazil’s UBC distributed more than BRL 1 billion to members in 2025, nearly 8 percent more than the prior year, and accounted for 62 percent of royalties distributed by ECAD. Those figures put a useful scale around rights administration. In AI debates, collective management is not abstract plumbing. It is an operating system for getting large sums to creators.
UBC’s chief executive credits technology and innovation while emphasizing creator protection, according to CISAC. That pairing matters for AI music because rights organizations cannot choose between modernizing their systems and defending repertoire. Their leverage depends on doing both well, with reliable data, transparent matching and distributions creators can audit. The record payout does not validate any particular AI policy, but it shows the institutional capacity already at stake.
