The BBC’s Director Of Music, Lorna Clarke, has outlined the broadcaster’s policy on AI in music, stating that the Beeb’s radio stations and other platforms will always prioritise music that is “the result of meaningful human creativity”, but that doesn’t mean those humans can’t use AI tools.
However, the BBC wants to be transparent about how AI has been used in the music it broadcasts and champions, which will require artists and labels to be transparent about the use of AI tools when submitting their music. Which will be a bit of a challenge, in the short term at least, as the industry itself hasn’t yet worked out how the use of AI in music creation should be identified, declared and labelled.
Clarke also says the BBC will “never knowingly broadcast AI-generated music that infringes existing copyright works”. Which may also be a challenging commitment to meet, depending on whether that means AI-generated music that clearly infringes existing works in the output, or any music generated or manipulated by an AI model that was trained on unlicensed music.
If it’s the latter, most AI companies are very vague about what existing content they have used to train their models. Initiatives like Ed Newton Rex’s Fairly Trained - which certifies AI models trained exclusively on licensed content - can help to an extent. However, as it currently stands, the chances are many of the AI tools being used by music creators were trained on unlicensed works.
As a crucial public service broadcaster, the BBC will always be subject to extra scrutiny when it comes to things like AI policies. Last year it came under fire when the West Midlands edition of ‘BBC Introducing’ included a pretty mediocre AI-generated track in its Artist Of The Month spot. Even though it felt like it only did so in order to provide a reason to talk about music AI tools on air.
In a blog post published yesterday, Clarke says “as the role of AI in music evolves, it’s important to reflect on how the BBC approaches AI-generated music content across our radio stations and platforms. Our starting point is clear: the BBC’s AI principles - prioritise talent and creativity, be open and transparent and we’ll act in the best interests of the public”.
After noting that musicians have long employed new technologies in their music-making - and that now includes AI tools - she writes, “artists using AI tools will be considered in our programming, as we expect more to experiment with and adopt this technology over time”.
However, she clarifies, “these tools must be used to support human creativity and should not replace a person’s work in developing, shaping or expressing the creative idea itself”. Which means, Clarke says, “simply prompting, generating, selecting or lightly modifying AI outputs is unlikely to represent meaningful human creativity”.
The BBC’s transparency commitment is informed by audience research, Clarke explains. “Audiences care about musicians and human creativity and may want to know when AI has been used to support the creation of the music they are listening to”.
To that end the BBC wants to flag uses of AI in the music it plays, and “to do this, we now require artists and partners submitting their work to us be transparent about whether and how they’ve used AI in that work”.
This feeds into an increasingly important debate within the music industry regarding when and how the use of AI in human created music is documented and declared, between collaborators, between artists and their labels and distributors, and with streaming services, media platforms and other users of music.
Opinion is divided on quite how effective automatic tools for detecting AI music are, especially once we are talking about AI-assisted music rather than tracks entirely generated by an AI model. But there aren’t really yet any consistent rules across the industry regarding self-declaration of AI use, and the need for such rules is arguably getting urgent now.