IFPI adopted global principles on Thursday governing whether recordings made with generative AI can appear in official music charts, and began applying them across the chart programmes it directly runs. The rules were proposed the previous day by the three majors and a group of independents.

The six criteria

A recording qualifies only if: the AI service used was properly authorised and lawful; the track is "substantially human made"; there are no stream or chart manipulation concerns; it complies with copyright and personality rights; it does not breach the AI service's terms; and AI use is signalled to consumers on streaming services. Signatories to the proposal: Sony Music, Universal, Warner, Believe, BMG, Concord, Dirty Hit, Glassnote, HYBE, Mom+Pop and Partisan.

"Worldwide" is the weakest word in the announcement

IFPI binds the charts it directly manages. Live now: the Official MENA Charts, the Official Southeast Asia Charts, the Official South Africa Chart and several Latin American charts. A pipeline of more than 20 runs through national groups including ARIA in Australia, SNEP in France, the Offizielle Deutsche Charts and Circle Chart in South Korea — described as in progress, not adopted. Billboard and the UK Official Charts Company are not covered and have committed to nothing. "AI songs banned from the charts worldwide" is false.

The threshold is stricter than "fully AI-generated"

Under the definitions drawn from RIAA and IFPI labelling work, AI-generated means made entirely by AI or with an AI-produced lead vocal or key instrumental. A track written, performed and produced by humans but with an AI-generated lead vocal is disqualified. AI-assisted — made mostly by people using AI for some elements — remains eligible. There is no published effective date, no audit mechanism, and enforcement rests on a soft "reason to believe" standard.

Who wrote the rules

These principles were drafted by rightsholders and adopted by their trade body — the parties with the most direct commercial interest in which records count. That is not a reason to dismiss them, but it should be stated rather than presented as a neutral standards body acting alone. Note also the first criterion: requiring that the AI service be "properly authorised and lawful" imports the training-licence question straight into chart eligibility — one day before a Munich court ruled on exactly that question.