THE MUSIC MANAGER’S

INTERIM

GUIDE TO AI

SONGWRITER

DISSECTING THE
DIGITAL DOLLAR
themmf.net/digitaldollar
This is a guide for music managers on how
artificial intelligence is impacting on the
music industry, helping them to navigate and
understand the opportunities and challenges
posed by AI, and offering guidance on how
they might advise their clients.

It looks at how the music community is using

– and will use – AI tools, and also considers
some of the potential threats, alongside how
music-makers can safeguard their rights
as ever more sophisticated AI models are
developed and employed.

We have called it an ‘interim guide’ because

both the technology and the regulatory
framework are evolving so fast it is going
to need regular updating. Please check
back on the MMF website to ensure you are
reading the most recent version.

THE MUSIC MANAGER’S (INTERIM) GUIDE TO AI was written by

Chris Cooke and Sam Taylor from CMU, with input and expertise from
Simkins, Russells, Level, Redburn Atlantic, Blackstar, and the MMF
board, team and membership. Published September 2023.
Section One: What is AI?
Artificial intelligence factory robot is flexible, accurate
and consistent, but not intelligent.
– while not new – has
Much research has humans program
clearly become a much machines to behave in a clever way,
bigger talking point like playing chess, but, today, we
in the last year. Partly emphasise machines that can learn, at
because of developments least somewhat like human beings do”.
in the evolution of AI Depending on how you define
technologies. And partly artificial intelligence, certain
because governments and technologies commonly referred
law-makers around the to as AI may or may not fit that
world have been giving definition. Some technologies – while
employing clever algorithms and a
more consideration as to certain level of machine learning –
how they might regulate may not be truly AI.

T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I
different AI models and
platforms. But whatever definitions we employ,
what is certain is that AI technologies
AI is also starting to have a much will be able to perform ever more
bigger impact on the music industry, sophisticated tasks that involve
with AI tools increasingly being used something very similar to human
as part of the music-making process, intelligence, and that this will have
and to assist with music marketing a big impact on the economy and
and other music business activities. society at large, way beyond the
music industry.
Meanwhile, rights-holders like record
labels and music publishers are It is anticipated that certain jobs –
starting to enter into deals with AI including creative and administrative
companies. roles – could ultimately be replaced
by AI. And there are also concerns
In terms of defining ‘artificial about how AI models might be used,
intelligence’, a term coined by a and the impact that could have on
Stanford University professor in 1955, privacy, security and democracy.
here is what Stanford’s Institute For
Human-Centered AI has to say… At the same time, AI creates
significant economic, commercial
“Intelligence might be defined as the and creative opportunities, though
ability to learn and perform suitable there is general agreement that
techniques to solve problems and competent regulation will be
achieve goals, appropriate to the required to ensure that the positives
outweigh the negatives in the short
context in an uncertain, ever-varying
and longer term.
3
world … a fully pre-programmed
 GENERATIVE AI In terms of generative AI and
music, it is also worth distinguishing
In the context of the music industry –
between…
and the wider creative industries – of
particular interest is generative AI, a
n Those AI tools that can assist
specific kind of artificial intelligence.
human beings in the music-making
This term refers to AI models that are
process.
able to generate original content,
whether that is text, image, audio or
n Those AI tools – or similar – that
video, and, of course, music
generate music by stitching together
pre-existing musical segments or
Again, generative AI is not particularly
stems based on criteria set by a user.
new, but it has become a much
bigger talking point this year because
n Those AI tools – or similar – which
certain generative AI models have
help people to generate tracks that
become a lot more sophisticated,
imitate the style, sound or voice of
especially in the text and image
specific artists.
domain.
n Those AI models that actually
Meanwhile, a number of high-
compose and produce original music.
T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I

profile generative AI platforms have

become widely and easily accessible,
As of summer 2023, the latter group
meaning many more people are now
of AI models – which would include
interacting with generative AI models
Meta’s MusicGen and Google’s Music
and seeing what they can achieve.
LM – are still quite restricted in terms
of the length and quality of the music
As with artificial intelligence at large,
they can generate.
how you choose to define generative
AI will have an impact on what
It’s not yet clear how long it will
tools and platforms fall under that
take to overcome these restrictions.
banner. There may be some tools
Though given the speed with which
and platforms commonly referred to
image generation AI has evolved
as generative AI which don’t strictly
in the last year, it could happen
meet the definition.
relatively quickly.

4
 music

sound time

pitch energy rhythm

frequency temperature volume timbre duration alternation

Generating music with AI often involves using machine learning

algorithms to analyse large datasets of existing music –
the algorithm examines patterns and structures in the music,
breaking down each track into its constituent parts,
and using this information to generate new music.
 Section Two: Using AI
There are many ways in a huge amount of content around
their releases and marketing
which the music industry
campaigns, and to keep fans
is already employing AI engaged on social media platforms,
– including generative including written content, as well as
AI – as part of the music- visuals and video. An assortment of
making process; to help AI tools are now available to help with
this process, sometimes generating
generate and distribute content from scratch, sometimes
marketing content; to reformatting and repurposing content
better manage rights and for different channels and uses.
royalties; to handle basic
legal and admin tasks; and Data management
AI is increasingly being used in the
to more effectively run music industry to help with data
music businesses. management, both the rights and
metadata that is a key part of music
T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I

And artists and managers operating distribution and rights and royalty
at all levels are now starting to make management, and the huge amount
use of AI tools. That includes… of consumption and fan data that
flows back into the music industry
from digital platforms. AI can be
Songwriting and recording
used to identify issues in rights data
Music-makers are increasingly using
that are impacting on payments and
generative AI tools – or similar – to
can also help the industry identify
assist in the music-making process,
learnings from fan data.
either to help with ideas, or to generate
some elements of a new work.
Document creation
Production and mastering All companies, including music
AI tools can be used to help with the companies, are starting to use AI
production process, and especially to generate first drafts of business
the mastering process, when creating documents, including marketing
new recordings. There is much materials, but also logistical and even
debate regarding the ability of AI to legal documents. AI can also be used
master recordings compared to hiring to edit down or pull key information
a mastering engineer, although where out of documents, or to compare
independent artists are working on the content of different documents.
very tight budgets, AI tools can be Although some generative AI tools
very helpful. are very impressive, you should
always be aware of the limitations
and generally use these tools to
Marketing and visual content
6 Music-makers now need to generate
create initial drafts and always read
through carefully before publishing or Grammarly: A popular tool that uses
sending on. an AI-enabled algorithm to review the
spelling, grammar, punctuation and
Translation clarity of text.
AI tools that translate copy from
one language to another are Otter.ai: An AI-powered transcription
becoming ever more sophisticated. tool, which is very good at generating
There are various potential uses transcriptions of recordings of
for this in music. That might include meetings and conversations,
automatically creating versions of particularly where there are multiple
tracks in other languages, or allowing speakers.
fans to automatically access social
DreamStudio from Stability AI: This
media posts and other marketing
is good and straight forward prompt-
content in their language of choice.
based image generation tool. After
Again you should probably check
some practice, you can generate
automated translations before
great images with very little effort.
publication to make sure there are no
glaring translation errors which could Midjourney: Another popular AI tool
embarrass your client. for generating images from user-

T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I
prompts.
Initial research and ideas
Generative AI can also be used to Kaiber: This is a creative platform
help with research and ideas as that uses AI to generate videos and
part of projects, including ideas for images based on user-prompts.
marketing campaigns and general
Gen 1 / Gen 2 by Runway: More
business development.
useful AI tools for video creation.

AI TOOLS TO CONSIDER Voice-Swap: An AI tool designed by

We surveyed a few music managers DJ Fresh and Nico Pellerin that can
and marketeers about the AI tools be used to transform vocals so that
they are currently using, and here they sound like one of a number of
are ten that they suggested that featured artists who have partnered
managers might want to check out... with the company.

ChatGPT: A lot of managers and Elevenlabs: This a great tool for AI-
marketers have been using ChatGPT generated voice and voice cloning.
to generate initial drafts of marketing
These are, of course, just a small
or other copy. It has many other uses
number of the AI tools now
too, especially if you make use of
available. MMF will be compiling and
ChatGPT plugins.
maintaining a fuller list, so please
Bard: The generative AI tool from do let us know about any that you
Google that can also be used to have found useful. And do also read
generate text from prompts. our guidance on things to consider
before using AI tools on page 17. 7
 Section Three: Licensing AI
TRAINING GENERATIVE AI scenarios where people can make
use of copyright protected works
Music-making generative AI is
without securing a licence.
‘trained’ by being exposed to
human-created music. This may be
That commonly includes things like
music specifically created for training
private copying, critical analysis,
purposes, or from production music
parody and so on.
libraries, or commercially released
recordings and songs.
In most countries there are no
exceptions that seem to apply to the
Where a tech company trains a
training of a commercial AI model,
generative AI model with existing
although that is possibly still to be
music owned by third parties, it will
tested in court, for example regarding
make a copy of that music onto its
fair use in the US.
servers. This poses a number of
important copyright questions for the
And some tech companies argue
whole music community.
that if there is a relevant exception in
T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I

just one jurisdiction, providing they

Does the company need a licence to
base their servers there, they do not
make those copies from the relevant
require a licence.
copyright owners and, assuming
it does, how will that licence work
The music industry – and other
and how will any income generated
copyright industries – are now
be shared between the different
lobbying law-makers in many
stakeholders in the music industry?
countries seeking clarity regarding
the copyright obligations of AI
RIGHTS-HOLDER companies.
CONSENT
Does an AI company need to secure If it does transpire that the training of
consent and negotiate a licence with AI can be undertaken without licence
whoever owns the copyright in the in a small number of jurisdictions,
music used to train its model? The the copyright industries will lobby
music industry says a firm “yes”. policy-makers in all other countries to
interpret or amend copyright law in a
However, some AI companies have way that counters that fact.
argued that the training of AI models
is covered by a copyright exception, That would mean that models trained
usually an exception relating to data with unlicensed works relying on a
mining or possibly the more generic copyright exception could not be
concept in US law of ‘fair use’. commercially exploited in countries
where no such exception exists.
8 Each copyright system does provide
a number of copyright exceptions, In practical terms, that would mean
that platforms using those unlicensed to distinguish AI-generated works
AI models would have to be geo- from human-created works, and
blocked in those countries. Internet would also allow rights-holders to
service providers could also be asked identify which works have been
or forced to block access to these generated by any one AI model.
services and search engines to de-list
them. Second, each AI company should
clearly declare what material
TRANSPARENCY + was used to train its models, with
an obligation to keep and make
LABELLING
available a complete record of all
Assuming it is agreed that the training works that were used as part of the
of generative AI with copyright training process.
protected works requires a licence,
how do copyright owners know if
their works have been used?
LABEL AND PUBLISHER-
LED DEAL-MAKING
In theory it should be possible for Certain players in the music
an AI company to keep a detailed industry – ie labels, distributors and

T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I
record of every work used to train publishers, and especially the major
any one AI model, and to identify music rights companies – are already
which specific works provided negotiating deals with certain AI
learning for and/or influence over any companies, which will allow those AI
subsequently generated content. firms to train their models using some
or all of the existing music owned or
Whether this is happening is another controlled by the music businesses.
matter. Some AI companies may not
store this data, either for efficiency It is not yet clear how those deals
reasons or to avoid future liabilities. will work. Obviously, the nature of
Some AI models will also build upon each deal will depend on how an AI
learnings from other AI models, and company plans to specifically make
while in theory a complete record of use of the existing music, and how
all works used for training should still their AI tools will subsequently be
be possible, data is more likely to be employed and monetised.
lost in that scenario.
As with the licensing deals that have
The music industry – and other been negotiated in the streaming
copyright industries – are also market, music-makers can be at a
lobbying for AI companies to disadvantage here. They and their
have wide-ranging transparency managers are one step removed from
obligations in law. There would be the deal-making process, deals being
two elements to this: negotiated by rights-holders and
platforms.
First, that content generated using
generative AI should be clearly And they are not usually allowed 9
labelled. This would allow consumers to see the specific terms of any
 one deal, making it hard for them the song copyright were exploited.
to evaluate if it is broadly fair and if PPL would have to be involved if
payments are correct. there was a case for arguing that
a performer’s right to equitable
However, there are a number of remuneration applied.
questions music-makers and their
managers might want to ask any However, the collecting societies
labels, distributors or publishers that may have a wider role to play than
they work with. this, depending on the technicalities
of how each AI model works and the
n Will rights-holders license their complexities noted above around
full catalogues – or large portions of identifying how any income should
their catalogues – for the training of be allocated to any one recording or
generative AI, or will they work with song.
AI companies on an artist-by-artist,
writer-by-writer, work-by-work basis? If the music industry is sharing in any
income associated with each new
n Will deals involve an upfront and/or track generated by AI, how is that
ongoing training fee, and/or a share income allocated to individual works
of any revenue generated by the AI and rights-holders?
T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I

platform?
How easy is it to identify which
n How will a rights-holder allocate existing works had a generic or direct
income to specific recordings or influence on the new work? And
songs – will an AI platform be able how is revenue sharing managed
and/or willing to provide data to if thousands or millions of existing
assist in this process? If not, how will tracks had an influence?
money be allocated so that income
can be shared with music-makers? One way to address these
complexities is to have a blanket
n What share of income allocated licensing system – or even a levy
to any one recording or song will system similar to the private copy
be paid through to the artist or levy that exists in some countries –
songwriter – ie what royalty rate will which is managed by the collecting
be applied? societies, with monies distributed
directly to writers and performers
THE ROLE OF according to rules set by each
society.
COLLECTING SOCIETIES
It is not yet clear what role, if any, the There are pros and cons to this
music industry’s collecting societies approach – and labels and publishers
will play. are likely to oppose going this route
– but it may be something music-
With UK repertoire, PRS would have makers want to consider.
to be involved if, at any point in the
10 process, the ‘performing rights’ of
MUSIC-MAKER CONSENT and publishers intend to seek the
specific consent of music-makers?
It is common in the music industry for
artists and songwriters to assign or
Some independent music companies
license the rights in their recordings
have explicitly stated that they will.
or songs to business partners.
Others have implied that they will,
including Universal Music when
Under conventional record and
announcing its involvement in the
publishing deals, the label or
YouTube Music AI Incubator in August
publisher owns the copyrights in the
2023. But it remains unclear what the
music (with songs, technically the
general industry position is on this.
publisher only owns those elements
of the copyright not assigned by the
However – oblivious of what
writer to their collecting society).
position any one label, distributor or
publisher might take – is there a legal
Even where artists and writers have
requirement for music-maker consent
retained ownership of their rights, if
to be sought?
they have entered into distribution
or administration agreements with
Contractual Veto rights

T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I
a label, distributor or publisher, they
may have granted their business Where artists and songwriters assign
partner wide-ranging control over the copyrights in their works to
their music. business partners, such as labels and
publishers, they will usually retain
Where this is the case, the label, some contractual rights.
distributor or publisher might be
empowered to license any music that This often includes the right to
it controls to an AI company without be consulted about and/or veto
seeking the specific consent of the certain uses of their music, such
music-maker, providing they pay as synchronisation, samples, or
any royalties the music-maker is due adaptations that will result in the
under contract. creation of a ‘derivative work’.

However, most music-makers and While these rights will vary

their managers would argue that from contract to contract – and
– given the novelty and potential conventions are different for record
impact of generative AI – specific deals and publishing deals – are
consent should be sought from each commonly included consultation or
individual music-maker before their veto rights relevant to the use of
music is used to train an AI model. music by AI companies?

The music industry at large has talked Moral rights

a lot about the need for AI companies The moral rights provided for creators
to seek consent before exploiting by copyright law vary greatly from
existing works as part of the training country to country, but usually include
process. But do labels, distributors the right to attribution and the right to 11
 object to the derogatory treatment of Publicity rights
any one work. Publicity rights are a legal concept
that allow individuals to control
Are any moral rights being impacted the commercial use of their image
upon by the training of generative AI or personality – indeed they
and, if so, does that mean the specific are sometimes called image or
consent of the creator is required? personality rights.

It is worth noting that moral rights are Quite how these work differs
relatively weak under UK law and can greatly around the world, and
be waived in record and publishing – as these rights have become
contracts. more commercialised in recent
decades – different sectors within
Performer rights the wider entertainment and sports
Copyright law provides performers industries have developed different
with certain rights over recordings on conventions for how they are
which they appear even when they managed and licensed.
are not the copyright owner.
In football, where the publicity rights
The performer’s approval is of players are routinely exploited by
T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I

required to make a recording of things like video games and brand

their performance and to exploit the partnerships, consideration has
resulting copyright. Performers are been given to how a player’s deal
also due equitable remuneration with a club handles these rights.
via the collective licensing system Players might establish a separate
whenever the performance or company to control their publicity
communication controls of the sound rights, and their club would enter into
recording copyright are exploited. a separate deal with that company if
they wanted to pursue commercial
Have performers granted the opportunities that exploit those rights.
necessary approvals to train an AI
with a recording of their performance In music, artists would usually
and could any uses of AI necessitate manage their publicity rights in
the payment of equitable partnership with management.
remuneration?
Under a basic record deal, a label
Under the current system, the would have permission to exploit
necessary approvals have probably the artist’s image in order to market
been provided under most industry the records it is releasing, but would
standard record contracts and not usually be involved in any other
session musician agreements. commercialisation of these rights.
However, copyright law could be
amended to grant performers specific However, some labels have started to
approval or remuneration rights in seek wider involvement in an artist’s
12 this domain. publicity rights, usually as a result
of new digital opportunities, such Therefore, arguably an artist would
as an artist having a representation need to provide specific consent if
on gaming platforms like Fortnite or a recording of their voice is stored,
Roblox. distributed or exploited, for example
by an AI company.
In the context of generative AI,
publicity rights are most relevant A label would likely infer that that
when an AI model imitates the voice consent has been provided through
of a specific artist. But do publicity an artist’s record deal when it
rights come into play in other uses of comes to the standard storage and
generative AI? distribution of recordings. However,
it’s a stretch to assume such an
And how will publicity rights be inference in the context of using an
managed and licensed in this domain, artist’s voice to train an AI model.
and to what extent will an artist’s
business partners want a greater Why this matters
involvement in these rights? The legal requirement for specific
music-maker consent – if it exists – is
It is important to note that in the UK important for a number of reasons…

T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I
publicity rights do not currently exist
in law. n For the industry at large, it could
provide additional control over
The tort of passing off can sometimes recordings and songs even if AI
be employed to stop people companies are able to find and
commercially exploiting an artist’s exploit copyright exceptions for some
image or personality, though that of their uses of music.
only usually applies when the entity
doing the exploiting is misleading the n It would allow music-makers who
consumer into thinking they have the have assigned the copyright in their
endorsement of the artist. music to business partners to still opt-
out of AI licensing deals on a case-
Given publicity rights could become by-case basis, and/or to negotiate
important in the context of generative specific terms with their label,
AI – and other areas like the distributor or publisher regarding how
metaverse – the concept probably income from AI deals will be shared.
should be introduced in UK law.
n Some of these rights – for example
Data protection rights publicity rights – may also allow
Data protection rights – and/or other music-makers to negotiate their own
privacy rights – may well give an deals directly with AI companies
individual control over the storage, alongside any licensing deals struck
distribution and exploitation of by labels and publishers, providing
recordings of their voice or image. new revenue opportunities for the
artist.
13
 Section Four: Campaigns and
consultations on music and AI
There a number of forums been reviewing the impact of new
technologies on the creative sectors,
where the questions and
including AI. It published a report in
issues raised above are August 2023.
currently being discussed
and considered, some Congressional Hearings:
instigated by governments Committees in US Congress have
staged a number of hearings on
and law-makers, some by AI, including sessions specifically
the music industry, and focused on AI and copyright.
some by organisations
in other creative sectors. US Copyright Office Consultation:
They include… The US Copyright Office has
opened a consultation seeking input
on various copyright issues and
GOVERNMENT-LED
T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I

questions posed by AI.

EU AI Act: The European Union is

in the final stages of agreeing new MUSIC INDUSTRY-LED
regulations for artificial intelligence
that will go into effect across all The Human Artistry Campaign: is a
member states. Although the global campaign bringing together
act is not specifically focused on creators and copyright owners from
generative AI, it does have some multiple creative disciplines. It was
measures regulating it, including on initially spearheaded by the US
transparency and labelling. record industry but has since been
embraced globally and by multiple
IPO AI Working Group: The UK’s creative sectors.
Intellectual Property Office has
convened a working group of The UK Music AI Policy Paper: sets
AI companies and organisations out key objectives and concerns
from across the creator and about generative AI on behalf of
copyright sectors to inform the UK various UK music industry trade
government’s position on AI and organisations, including those that
copyright. UK Music and the Council represent labels, publishers, artists,
Of Music Makers sit on this group and musicians, songwriters, studio
MMF is involved with the debate. producers and managers.

CMS Select Committee Inquiry: The Council Of Music Makers

The Culture, Media & Sport Select AI Agenda: the Featured Artists
14 Committee in the UK Parliament has Coalition, Musicians’ Union, Ivors
Academy, Music Producers Guild or exploitation of your work – for
and Music Managers Forum – which instance, forbidding AI-rendered
are all members of UK Music – have translation, editing, cover design,
also identified and communicated a indexing, and audio recordings”.
number of questions for labels and
publishers to ensure that music- The society also recommends that
makers are fully informed about and authors should “check the terms and
properly consulted on generative AI privacy settings of cloud services you
business and licensing models as use to store or develop your work.
they are developed. Some services by default reserve the
right to analyse your text and images
for development purposes. While not
OTHER CREATIVE
always explicitly stated, this could
SECTORS include using your work to train AI
systems”.
Creators’ Rights Alliance AI Policy
Paper: The UK Creators’ Rights US Authors Guild AI Model
Alliance – which brings together Contract Clause: In March 2023,
organisations representing creators the US Authors Guild published

T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I
from many different artforms – has a new AI-specific model clause
published a policy paper on AI. which it recommended be inserted
into publishing and distribution
Like the UK Music policy paper, it agreements within the American
sets out key objectives and concerns books sector. It explicitly states that
about generative AI, and calls on the no consent has been granted for
UK government to clarify and, where the use of an author’s work for the
necessary, extend the copyright purposes of training AI.
and transparency obligations of AI
companies. The Guild also advised authors
and their agents to check existing
UK Society Of Authors AI Policy agreements for any terms that grant
Paper + Practical Guidance: The a publisher the right to use their
UK Society Of Authors has published work in relation to the training of
a policy position on AI as well as AI, or more generally for things like
practical guidance for its members. ‘internal purposes’, ‘research’ or ‘data
It recommends that authors include mining’. Where such terms are in an
explicit terms in new agreements agreement, it said authors and agents
to the effect that a publisher “may should seek to have them removed.
not use, or allow access to, the
work in any manner which could Equity AI Vision Statement:
help the learning/training of artificial UK performers union Equity has
intelligence technologies”. published an AI Vision Statement
which outlines eight principles
It added that: “You might want to that it believes the industry should
prevent AI technologies being used
in connection with the creation
adopt in relation to AI. It includes 15
 a commitment around performer “It is our position that language in a
consent. performer’s contract which attempts
to acquire the rights noted above are
It said that artists should “have the void and unenforceable until terms
right to consent (and not consent) have been negotiated with SAG-
when their performance, voice AFTRA. The rights have not been
or likeness is recorded, used or conveyed”.
reproduced by machine learning
systems and equivalent technology. One of the demands in the ongoing
This right cannot be assigned or SAG-AFTRA strike is that US studios
waived, and applies in perpetuity for and production companies must
past, current and future performances provide guarantees about how they
or likenesses”. intend to use AI and how the rights of
performers will be protected.
SAG-AFTRA AI Policy Position: In
June 2023, the US performers union LITIGATION
set out its position on AI in a letter. It
In addition to these campaigns
stated that: “The use of performer’s
and consultations, a number of
voice, likeness or performance to
lawsuits have been filed in multiple
train an artificial intelligence system
T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I

countries which will test the copyright

designed to generate new visual,
obligations of AI companies. To date
audio, or audiovisual content is a
these have mainly been pursued by
mandatory subject of bargaining”.
the book publishing and photography
sectors, with key cases including
“You cannot unilaterally impose terms
those filed by various authors against
in individual contracts that purport to
OpenAI’s Chat GPT and those filed
grant these rights. We are entitled to
by Getty Images against image-based
bargain over the compensation and
generative AI platform StabilityAI.
terms under which these rights are
granted and used”.

16
Section Five: Advice for managers
USING AI Copyright liabilities
When music-makers and managers Remember the ambiguities outlined
are making use of AI tools as part above regarding the copyright
of the music-making process, or obligations of AI companies and
to create marketing content, or the current lack of transparency
otherwise, there are a number of regarding what materials have been
things to consider… used to train any one AI model.

There is likely a lot of litigation ahead

Reliability
in this domain and some copyright
Although some generative AI tools
owners might target the users as well
are already very impressive, they are
as the owners of allegedly infringing
not without limitations, and these
AI tools and platforms. And this
need to be factored in.
may discourage some companies
– especially bigger companies and
For example, if you use generative
brands – from making use of these AI

T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I
AI to create marketing copy, you
tools in the short term.
should always see the AI-generated
content as a first draft and, when
For example, there have been reports
checking that content, be aware
that the big marketing agencies are
that sometimes an AI can generate
not yet using AI-powered image
factually incorrect statements.
generation tools for fears of future
liabilities, instead waiting for the
So, a certain amount of fact-checking
copyright obligations of those AI tools
is always advisable.
to be clarified by law-makers or in the
courts.
Granting of training rights
When making use of AI tools, be wary
of what rights you are granting the
Copyright status of
AI company via its terms of service. generated works
The AI model is likely being further There has been much debate
trained by the prompts you provide regarding the copyright status of
when using the tool and any existing AI-generated works – ie does AI
content you upload as part of the generated content enjoy copyright
prompting process. protection?

Make sure you are aware of any UK copyright law provides ownership
rights you are granting to the AI rules for ‘computer-generated works’,
company and – if you are working which suggests that those works
with content where you are not the enjoy copyright protection.
copyright owner – that you are in a
position to grant those rights. However, most copyright systems
are silent on this and it is assumed 17
 that AI-generated works do not enjoy to an AI company without getting
copyright protection. specific consent.

This means if you use generative n Irrespective of what any one

AI to create – for example – contract says, the MMF advises
artwork, that artwork may not be music-makers and their managers to
protected by copyright, which would formally inform all and any business
limit your ability to monetise the partners that no consent has been
content. Because, if the artwork granted for the use of each music-
was considered ‘public domain’ maker’s music for the training of AI.
in copyright terms, anybody could Even where consent is arguably not
make use of that artwork, including required under contract, some of
commercially, without getting your the other rights outlined above may
permission. be relevant, and the industry may as
yet more widely adopt the position
In the US, the Copyright Office has that music-maker consent is always
stated that AI-generated works do required.
not enjoy copyright protection but
AI-assisted works do. This obviously The UK’s Council Of Music Makers
poses a big question about how has published a template letter that
T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I

much creative input a human being music-makers and managers can

must have when using AI tools for a use to confirm to business partners
work to be considered AI-assisted that consent has not been granted
rather than AI-generated. and should not be assumed. That
template is on the next page and also
LICENSING AI – available via the CMM website.
MUSIC-MAKERS LOCKED
INTO EXISTING DEALS LICENSING AI –
Where music-makers are locked into
NEW DEALS
existing record or publishing deals When music-makers are negotiating
via which a label or publisher owns new deals with labels, distributors or
the copyright in the artist or writer’s publishers, consideration should be
music, or has wide-ranging control given to the future use of music in the
over the music, what should the AI domain.
music-maker and their manager be
doing? n Be very careful of granting
business partners any rights beyond
n It is worth reviewing any existing copyright, for example publicity,
contracts and identifying what image and personality rights. Where
specific rights have been granted labels seek involvement in these
to each business partner, what rights, be very clear and specific on
exclusions and veto rights are what that involvement is.
included, and whether that means
n Seek specific AI approval rights
18 the business partner can license the
in contract – so that any use of the
music-maker’s recordings or songs
This is a template letter provided by Council Of Music
Makers that music-makers and their managers can tailor
and send to any labels, distributors and publishers they
work with, making it clear that advance permissions must be
sought before any of their music is used to train an AI model.

Dear X

I am currently reading about the potential impact of artificial intelligence

– and especially generative AI – on the music community and the wider
creative industries.

It’s good to see that so many UK music companies and organisations

are embracing the Human Artistry Campaign. I agree that “copyright
protection exists to help incentivise and reward human creativity, skill,
labour and judgement”, and that “creators and copyright owners must
retain exclusive control over determining how their content is used” to

T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I
ensure “that human creators are paid for their work”.

It is important we acknowledge the potential positive impact of AI

on the music business, while also stressing that AI companies and
their business partners must respect copyright, and other creator and
personality rights, and secure consent from music-makers before
making use of their music.

For this to occur it is essential that tech companies are fully transparent
about how they train and utilise any AI models for music and, in turn,
that rights-holders are transparent about licensing deals covering these
models and the income generated from them.

For the record, I do not currently consent for any [recordings that I
performed on / songs I wrote or co-wrote] (delete as appropriate)
to be used to train any AI models. I also consider that the use of my
works in the context of the production of any derivative works to be an
unauthorised adaptation and an infringement of my moral rights.

I look forward to hearing about any opportunities in the music AI

domain that you identify and discussing how we might collaborate on
pursuing those opportunities together.

Many thanks
[INSERT NAME]

You can also access this template at councilmusicmakers.org/ai-letter

19
 artist’s music in the training of AI MMF has drafted a sample contract
requires explicit consent on a project- term to provide guidance on some of
by-project basis. the protections managers might seek
to include in new contracts in this
n Seek specific AI transparency domain.
obligations in contract – so that the
rights-holder is obliged to disclose This is intended to simply inform
to the music-maker when their work contract negotiations, and managers
has been used for AI training or for should consult the artist’s lawyer for
AI purposes more generally. That further guidance.
includes if the artist’s music or image
is being uploaded into an AI platform The sample contract term is as
in order to create marketing content. follows: “Explicit artist consent must
be secured before any music or other
n Ensure that any AI transparency content produced or created under
obligations extend to the AI this contract is used or licensed for
companies themselves – so the the training of generative AI models.
rights-holder commits in the artist’s The artist retains control over their
contract to ensure that transparency professional name, image, brand,
obligations are included in any logo, face, likeness, voice, signature
T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I

licensing deals with AI companies. and other identification attributes

– and any publicity, personality,
These kinds of deal terms are not image or similar rights provided in
yet standard in label, distributor and law. Explicit artist consent must be
publishing contracts, but they are secured before any of these rights
starting to be discussed as part of the are exploited in connection with any
deal-making process. AI model or platform.

Therefore managers and lawyers As the debate around music and AI

should always ensure that these is rapidly evolving, MMF will likely
things are considered in any deal- update its advice and guidance on
making and covered in any new a regular basis, so check the MMF
contracts. website for the latest updates.

20
Section Six: Next steps
Campaign with the music be allocated to individual works, and
industry for clarity on what royalty rates will be applied to
rights-holder consent and AI income.
transparency
Music-makers and their managers n A discussion on the role of
should work with the wider music collecting societies, including where
industry – and the wider copyright complexities around licensing and
industries – for example via the royalty distribution might be best
Human Artistry Campaign, to ensure dealt with via a collective licensing
that: approach.

n copyright obligations of AI Review and campaign for

companies are clarified and that stronger music-maker rights
no new copyright exceptions are The music-maker and management
introduced that can be exploited to communities should review the
circumvent these obligations. general legal requirements regarding

T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I
music-maker consent – including in
n clear transparency obligations are copyright law, but also as a result of
applied to generative AI models, so publicity rights and data protection
that AI-generated works are clearly law. This may require more clarity
labelled, and a detailed record is kept in law and/or an extension of rights,
and made available on what works including the introduction of publicity
have been used to train any one rights in UK law.
model.
Confirm position with
Campaign within the music existing business partners
industry on music-maker Music-makers and their managers
consent and deal transparency should confirm to any labels and
Music-makers and their managers publishers they work with that they
should campaign within the music have not provided their consent
industry – via organisations like the for their music to be used to train
Council Of Music Makers – in order generative AI models. While there
to seek: may remain some debate as to what
consent is required – if any – by a
n A commitment from rights-holders label or publisher, it is good to be
that specific consent should be clear that no consent should be
sought from each music-maker assumed or implied.
before any music is licensed to and
used by AI companies. Consider AI in all new deals
When music-makers and their
n Clarity on how AI licensing deals managers are negotiating new deals
will be structured, how monies will with labels and publishers, they 21
 AI CHECKLIST
Music-makers and managers should consider how
AI tools can enhance and assist their businesses

Music-makers and managers should be aware of what

rights they are granting when they use AI tools

Music-makers, managers and the wider music industry

should campaign for clarity on AI company obligations

AI companies must seek consent from rights-holders

and be fully transparent about training datasets

AI companies and rights-holders must

seek explicit music-maker consent

Rights-holders should involve music-makers and

managers in the development of AI business models

Rights-holders should be fully transparency

about their AI deals and revenues

Managers should review existing music-maker record

and publishing contracts in the context of AI

Managers should ensure that AI considerations are

included in new music-maker contracts
should ensure that music-maker They should also be careful when
consent is required before their granting business partners any rights
music is licensed for the purposes beyond copyright, such as publicity
of training AI models, and also seek rights, and by very clear what the
transparency obligations from each business partner’s involvement will
rights-holder regarding any use of the be in these areas.
music-maker’s music by AI.

T H E M U S I C M A N AG E R ’ S ( I N T E R I M ) G U I D E TO A I

23
This is an interim guide for music
managers on how artificial intelligence
is impacting on the music industry. It
has been produced by music consultancy
CMU for the Music Managers Forum.

ABOUT THE MMF ABOUT CMU

MMF is the world’s largest professional CMU helps people navigate and
community of music managers in the understand the music business through
world. Since our inception in 1992 we media, training, consultancy and events.
have worked hard to educate, inform and
represent our managers as well as offering Our media includes the CMU Daily bulletin,
a network through which managers can Setlist podcast and online CMU Library.
share experiences, opportunities and
information. We offer our own music business training
programme and deliver training courses for
We are a community of over 1500 music companies and organisations.
managers based in the UK with global
businesses and a wider network of over Our consultancy work sees us provide
2700 managers globally. We engage, strategic support, intelligence and
advise and lobby industry associates and expertise to a wide range of clients in the
provide a professional voice for wider UK, EU and around the world.
industry issues relevant to managers.
We also regularly produce and publish
The MMF runs training programmes, research reports and white papers on the
courses and events designed to educate very latest trends in the music business.
and inform artist managers as well
as regular seminars, open meetings, completemusicupdate.com
roundtables, discounts, workshops and
the Artist & Manager Awards.

themmf.net