Opinion

As streaming platforms battle an influx of music generated by Artificial Intelligence, Steve Harris talks to composer/bandleader Asha Parkinson about why she’s taken her debut album off Spotify

Technology is unstoppable, and legislation struggles to keep up. In 2025 there was outrage over a proposal to allow tech companies by default to use copyright material for AI training. Earlier in 2026, the Technology Secretary, Liz Kendall, said ‘we have listened’ and ‘it will take time to get this right’.

In 2024 Spotify CEO Daniel Ek joined Meta’s Mark Zuckerberg in telling European regulators to make things easier for open-source AI developers. He also said it was Spotify’s early adoption of AI that allowed it to provide ‘a personalised experience for every user’.

But when it comes to AI-generated music, Spotify has said it is strengthening its protections, noting that ‘Spam tactics such as mass uploads, duplicates, SEO hacks, artificially short track abuse, and other forms of slop have become easier to exploit’ due to simple-to-use AI tools.

The company’s new spam filter aims to identify AI tracks and stop them being recommended – it says that it has already removed more than 75 million of them. Meanwhile, Deezer reported that by November 2025 it was receiving 50,000 AI-generated tracks per day.

So, we’ve got more than enough AI music. I talked about this with composer and saxophonist Asha Parkinson [pictured above], one of the protesters who withdrew their music from the Spotify platform in the middle of last year [HFN May ’26].

One size doesn’t fit all

Asha’s multicultural musical background stems from a project she did in her teens, Voices Beyond Divisions, bringing together children from different faiths to perform her choral work What War in a concert that raised funds for the charity Syria Relief. For this she received a Diana Legacy Award in 2017.

Dismissive of AI ‘gunk’, Asha is also critical of the 1000-stream payment threshold imposed by Spotify and Deezer. ‘It is filtering out a lot of music, so that the most homogenised listener tastes dominate the platform. And there are a lot of algorithmic features on Spotify – like the Spotify Wrapped feature, the Discover Weekly, the Daylist and stuff, that “personalise” listeners’ tastes and make people feel quite seen. But actually, the things that get through to those algorithmic playlists have to have a lot of streams. People who are releasing jazz or classical music, or less mass-media approved genres, really struggle.’

In early 2026 Qobuz announced an AI Charter and ‘concrete steps to protect artists and listeners from AI-generated content’. Says Asha: ‘AI-generated artists like Velvet Sundown and Sienna Rose are still on Qobuz, but the emphasis of its Charter is that there wouldn’t be AI-generated playlists, etc – these would all be human-generated.

‘AI content would get naturally filtered out, which does seem like a good system. I guess it’s just going to be so hard for it to scrape it all off the system, because there’s so much of it. At least it’s not actively promoting AI-generated material.’

Producer Giles Martin has emphasised the difference between generative AI and the ‘restorative AI’ techniques he used when he extracted John Lennon’s vocal from a 1977 home demo cassette, put it together with George Harrison’s guitar from a 1995 studio session and added new tracks from Paul McCartney and Ringo Starr. ‘Now And Then’ went to No 1 in 2023. Nothing could be bigger than a Beatles reunion, even a virtual, AI-infused one, so it was worth the effort. And it does manage to sound human.