Kylie Minogue songs already used to train AI as tech firms ‘strong-arm’ Australian government, music body warns

Wonderful news from the Antipodes: Kylie Minogue has achieved a kind of immortality that no amount of Neighbours reruns could have secured. Her songs have already been fed to AI models, according to the publishers' group raising the alarm, which means the training happened while everyone was still holding meetings about whether training should happen. The industry's great achievement of the past three years has been to convert 'we are definitely going to have a consultation about this' into a functioning data pipeline. Kylie, presumably, found out the way the rest of us do: after the fact, from a press release.

The genuinely charming detail is the arithmetic of the lobbying. Dozens of countries were approached to water down their copyright laws; Australia, per the group, appears to be the only one still considering it. Dozens. One. That is not a trend, that is a queue of jurisdictions saying no and a single government holding the door open while pretending to check the guest list. The word 'outlier' is doing a lot of heavy lifting in the report, and the word 'strong-arm' is doing the rest — an accusation aimed at firms who evidently found that hard lobbying in Canberra is cheaper than licensing a back catalogue.

One can already picture the ministerial talking points: a sovereign copyright framework, a modernising economy, an 'AI-ready' nation, and absolutely no acknowledgement that 'AI-ready' is being defined by the people who would like to pay nothing for the material. No firm is named in the reporting, no specific bill clause is quoted, which is the standard shape of these stories — the pressure is described, the mechanism stays in the room. Meanwhile the recordings are already in the training set, so the consultation will now concern a fait accompli with a nice font. Australia's comparative advantage turns out to be a willingness to be the last country at the party still arguing about the door charge.