FREE AI music newsletter: https://www.aimusicpreneur.com
Part 2 of my interview with Daniel Bedingfield, recorded 27 April 2026. His answer on AI labeling is one sentence long: label music by how it was actually made, and make the labeling pay, because a system nobody profits from is a system nobody uses.
SIQA's 3-tier classification, in full:
https://www.aimusicpreneur.com/ai-mus...
The SIQA Verified Registry and the mid-year report:
https://www.aimusicpreneur.com/ai-mus...
Luminate's single flat AI Generated tag:
https://www.aimusicpreneur.com/ai-mus...
Spotify's AI Persona badge:
https://www.aimusicpreneur.com/ai-mus...
Suno v6, the first licensed model:
https://www.aimusicpreneur.com/ai-too...
Part 1, AI finished his 20-year song:
Daniel Bedingfield Sat on This Song for 20...
Part 3, replace the industry with AI: [Coming soon]
CHAPTERS
0:00 "It still is stealing, but I think they might be missing the point"
0:23 Suno v6, Universal's ElevenLabs deal, and the lawsuits still running
1:22 Should Spotify and Apple Music label AI tracks?
1:41 Why labeling only works if labeling pays
2:14 The 4 lanes Daniel Bedingfield describes for AI music
2:42 Imogen Heap, AI-human hybrid, and 15 years of fan-owned music
3:37 The identity layer: one node every AI model has to pay through
4:55 Suno vs Udio, and the attribution layer that pays something back
5:27 "My best friends won't use AI because it is stealing"
5:54 Letting every artist choose how much of their catalogue trains a model
6:21 "The tech bros are not musicians": the AI and Music Institute in the UK
7:19 Universal and Sony's Suno talks stalled. Who gets paid when a deal lands?
7:55 Why Daniel has never seen a royalty since he signed
8:49 Publishing pays 80%. Recording pays 20%.
9:13 Eleven Music trained on Kobalt and Merlin, and why Mozart is a dark horse
9:44 My answer: SIQA's 3 tiers, the Verified Registry, Luminate and Spotify
-----
He describes 4 lanes running from fully AI-generated to fully human, and puts Imogen Heap in the hybrid lane for cloning her own voice as ai.mogen. He then describes the identity layer Imogen is building, where anything that trains on an artist's data, face, lyrics or voice has to pass through one node and pay before it can go anywhere.
On tools, he uses Suno, says Udio was better until it stopped working for him, and notes that Suno's attribution layer now pays something back. He wants artists to choose how much of their own catalogue is allowed into a training set, and he recommends the AI and Music Institute in the UK for anyone who wants a say in how that gets built.
Then he gets blunt about money. He has never seen a royalty since he signed, because his label kept recoupment ahead of income for 2 decades, and he says that is by design rather than bad luck. He explains why publishing pays a songwriter roughly 80% where a recording deal pays roughly 20%.
He points to ElevenLabs' Eleven Music, trained on Kobalt and Merlin catalogues, as a model that could send money back to independents, and calls Mozart a dark horse after meeting the team.
In my outro I cover SIQA's 3 tiers: AI-Assisted, Human + AI Hybrid, and Fully AI-Generated. I explain the SIQA Verified Registry, where an artist declares AI use at the source before release and anyone can look it up at registry.thesiqa.com or scan a track with the iOS app. I set that against Luminate's single flat AI Generated tag and Spotify's AI Persona badge, which tells you an artist's identity might be AI and nothing about the music.
Disclosure, and I say it on camera at 10:23: I sit on SIQA's Creative Advisory Council. I report on SIQA and I advise it. I do not run it.
Part 1 is the song AI finished for him. Part 3 is what happens to labels and managers.
#aimusic #sunoai #musicnews
FREE AI music newsletter: https://www.aimusicpreneur.com
Part 2 of my interview with Daniel Bedingfield, recorded 27 April 2026. His answer on AI labeling is one sentence long: label music by how it was actually made, and make the labeling pay, because a system nobody profits from is a system nobody uses.
SIQA's 3-tier classification, in full:
https://www.aimusicpreneur.com/ai-mus...
The SIQA Verified Registry and the mid-year report:
https://www.aimusicpreneur.com/ai-mus...
Luminate's single flat AI Generated tag:
https://www.aimusicpreneur.com/ai-mus...
Spotify's AI Persona badge:
https://www.aimusicpreneur.com/ai-mus...
Suno v6, the first licensed model:
https://www.aimusicpreneur.com/ai-too...
Part 1, AI finished his 20-year song:
Daniel Bedingfield Sat on This Song for 20...
Part 3, replace the industry with AI: [Coming soon]
CHAPTERS
0:00 "It still is stealing, but I think they might be missing the point"
0:23 Suno v6, Universal's ElevenLabs deal, and the lawsuits still running
1:22 Should Spotify and Apple Music label AI tracks?
1:41 Why labeling only works if labeling pays
2:14 The 4 lanes Daniel Bedingfield describes for AI music
2:42 Imogen Heap, AI-human hybrid, and 15 years of fan-owned music
3:37 The identity layer: one node every AI model has to pay through
4:55 Suno vs Udio, and the attribution layer that pays something back
5:27 "My best friends won't use AI because it is stealing"
5:54 Letting every artist choose how much of their catalogue trains a model
6:21 "The tech bros are not musicians": the AI and Music Institute in the UK
7:19 Universal and Sony's Suno talks stalled. Who gets paid when a deal lands?
7:55 Why Daniel has never seen a royalty since he signed
8:49 Publishing pays 80%. Recording pays 20%.
9:13 Eleven Music trained on Kobalt and Merlin, and why Mozart is a dark horse
9:44 My answer: SIQA's 3 tiers, the Verified Registry, Luminate and Spotify
-----
He describes 4 lanes running from fully AI-generated to fully human, and puts Imogen Heap in the hybrid lane for cloning her own voice as ai.mogen. He then describes the identity layer Imogen is building, where anything that trains on an artist's data, face, lyrics or voice has to pass through one node and pay before it can go anywhere.
On tools, he uses Suno, says Udio was better until it stopped working for him, and notes that Suno's attribution layer now pays something back. He wants artists to choose how much of their own catalogue is allowed into a training set, and he recommends the AI and Music Institute in the UK for anyone who wants a say in how that gets built.
Then he gets blunt about money. He has never seen a royalty since he signed, because his label kept recoupment ahead of income for 2 decades, and he says that is by design rather than bad luck. He explains why publishing pays a songwriter roughly 80% where a recording deal pays roughly 20%.
He points to ElevenLabs' Eleven Music, trained on Kobalt and Merlin catalogues, as a model that could send money back to independents, and calls Mozart a dark horse after meeting the team.
In my outro I cover SIQA's 3 tiers: AI-Assisted, Human + AI Hybrid, and Fully AI-Generated. I explain the SIQA Verified Registry, where an artist declares AI use at the source before release and anyone can look it up at registry.thesiqa.com or scan a track with the iOS app. I set that against Luminate's single flat AI Generated tag and Spotify's AI Persona badge, which tells you an artist's identity might be AI and nothing about the music.
Disclosure, and I say it on camera at 10:23: I sit on SIQA's Creative Advisory Council. I report on SIQA and I advise it. I do not run it.
Part 1 is the song AI finished for him. Part 3 is what happens to labels and managers.
#aimusic #sunoai #musicnews