Major Labels Oppose Suno ‘Training Number’ Redaction Request

Photo Credit: Possessed Photography

The public deserves to know how many recordings Suno used to train its models – at least according to the non-Warner majors, which are pushing back against Suno’s demand to keep the much-discussed tally and different data under wraps.

Universal Music and Sony Music made that position clear in a recent filing, days after Suno doubled down on its redactions demand. We’ve covered the sub-dispute in detail, but the short version is that Suno (as well as its Udio rival) is aggressively attempting to stop the figure – identified by the plaintiffs during a discovery marathon – from being publicly disclosed.

That refers specifically to the number of tracks used to train Suno’s models. The way the well-funded AI company sees things, divulging the total would benefit competing music-generation businesses.

At the top level – and as Universal Music and Sony Music emphasized in their latest filing – this is perhaps a curious position because Suno’s already acknowledged training on “tens of millions” of works.

But the precise count’s going public probably wouldn’t help the company from the PR or legal perspectives, and there’s always the possibility that the figure is even larger than many are anticipating.

Back to the non-Warner majors’ opposition, the plaintiffs are technically looking to prevent the redaction of two numbers.

First, after leveraging Audible Magic’s audio-fingerprinting technology to determine which of their recordings were allegedly ingested by Suno, Universal Music and Sony Music maintain that the resulting “Audible Magic number” should be given the unredacted treatment as well.

“Importantly, the Audible Magic Number provides necessary context for understanding the scale of Suno’s infringement of Plaintiffs’ works, and why Plaintiffs have asserted claims for only a fraction of the Audible Magic Number,” Universal Music and Sony Music wrote. “There is no legitimate basis (and Suno has not articulated one) that would justify concealing it from public access.”

As for the core training number, the “figure is necessary to understand the full extent of Suno’s copying—which bears on Suno’s fair use defense,” according to the plaintiffs.

Furthermore, on the public-knowledge side, the larger number is said to demonstrate “that Suno’s sweeping disregard for copyright ownership likely extends well beyond Plaintiffs’ own sound recordings,” and the wider “public is entitled to understand the scope of that conduct in full,” per the filing.

What about the aforementioned competitive harm concerns? The non-Warner majors downplayed the risk of commercial consequences as “at most, marginal” – in part by reiterating Suno’s own acknowledgement of an expansive training process.

“Having publicly acknowledged copying of that magnitude, Suno cannot credibly claim that disclosing a slightly more precise count of the audio files that its AI model trained on would reveal anything competitively sensitive,” the plaintiffs summed up.

With that, all eyes are on the judge’s decision – especially because the above-noted Udio, repped by the same legal team as Suno, is also working to keep a lid on its training number.