Separate complaints against Suno accuse it of using protected music without permission and allowing users to imitate a voice. Those allegations put two decisions in front of a catalogue owner: what music may enter an artificial intelligence (AI) service, and what a user may produce with it.
Separate the input from the output
A service’s access to music and a user’s ability to generate a vocal imitation are different uses. Approval of one should not silently answer the other. A catalogue owner considering an AI agreement needs to identify which music the service may use, what users may generate, and how each permitted use will be paid. That sequence matters because the party controlling a recording or composition may be reviewing a different interest from the person whose voice an output imitates.
Start with the proposed input. Ask the service to name the recordings or compositions it wants to use and the activities for which it wants them. Then review the output separately: can a user request a voice imitation, and does the agreement permit the service to provide it? The Suno complaints concern both protected music and user generated voice imitation, so a single approval labelled AI use would leave the owner with too little information to assess either activity.
This separation also makes a refusal usable. An owner might approve access to specified music while declining a particular output capability. Conversely, permission to create an output should not be read as permission to use every recording or composition associated with its style. Put the permitted music and permitted outputs in distinct terms so that a later question can be traced to the decision that actually covered it.
Follow the payment to the use
Revenue terms need the same separation. For every use the agreement allows, the owner should be able to identify the event that earns payment and the party that receives it. An AI deal must account for how each permitted use will be paid. A payment for access to catalogue material does not, by its wording alone, tell a performer whether imitation of their voice was approved or compensated.
The practical review is a mapping exercise. Place each proposed use beside its permission and payment term. If a service wants music for its system and also wants users to generate vocal outputs, ask for a payment answer for each. A credit field cannot fill an empty permission or payment field. A musicians’ union has called for consent, credit and remuneration in AI use; those are three questions to record, even when the same person ultimately answers them.
For an independent label, this means checking its own authority before it offers a broad grant. Catalogue control establishes what the label can discuss, but the agreement should identify whose approval covers each proposed use. If the label cannot identify the person who can approve a voice imitation, it should leave that capability outside its grant until the approval path is clear. That is a contract decision the label can make without assigning a value to a right it has not cleared.
Give output controls an owner
Once the parties define the permitted outputs, someone must be responsible for keeping the service within them. Ask how a prohibited voice request is handled, who reviews a complaint about an output, and what record the service will return to the rights holder. Those questions follow directly from the allegation that users could imitate a voice through Suno: permission written between businesses is only useful if user activity can be checked against it.
A workable record should connect the disputed output to the relevant permission. The rights holder needs enough information to determine whether the output fell within the agreed scope, whether the appropriate approval existed, and whether a payment was due. The agreement should assign responsibility for producing that record. Otherwise, an owner can receive a complaint but have no way to decide whether it concerns an authorised use, an unpaid use, or a use the service was never allowed to offer.
Calls for studios to account for AI use, alongside musicians’ demands for consent, credit and remuneration, point to the same operational requirement: the business using the material must be able to explain what it did. For a catalogue owner, an explanation should connect an identifiable use to an identifiable approval and payment term. General assurances about responsible use do not perform that accounting task.
Use a two gate approval
Before approving an AI proposal, write down two decisions. First, specify the music the service may access and the payment attached to that access. Second, specify the outputs users may create, who can authorise them, and the payment attached to each allowed output. The scope of music use, generated content and compensation should be explicit in the deal.
- Catalogue gate: List the recordings or compositions covered and the service activity approved for each.
- Output gate: State whether voice imitation is allowed, whose consent covers it, and how an owner can challenge an output.
- Payment record: Match each approved use to its recipient, payment term and supporting usage record.
That gives an independent artist or label a concrete rule for its catalogue: do not treat permission to use music as permission to imitate a voice. Review and record both decisions before the service receives either approval.
