.png)
The Blue Highway Beacon: Vol. II, NO. 7
- Blue Highway Advisory
- May 7
- 5 min read
Updated: 1 day ago
How Can We Reward the Talent when their Creations are Used in AI-Generated Works?
By Amanda Harcourt, Honorary Professor of Practice, IBIL, Faculty of Laws, University College London, and Head of International Copyright, Contracts and Rights Administration for Blue Highway
Shall we call out moral panic or summon the pearl-clutchers?
In music, film, writing and art, Artificial Intelligence (AI) and its kissing cousin Generative AI invite conflict and concern.
Authors and performers rightly see that the powers in Silicon Valley, having already decimated their incomes over these last two or three decades, seem intent on a reprise. From the talent’s perspective, when it comes to AI and the training of this clever coding, those same Valley titans are proceeding based on the old principle of forgiveness rather than permission.
Same old swindle, and the talent has a point.
The creative industries have a problem to solve. Well, several actually.
Is the training of AI an infringement of copyright by copying? Google seems not to think so.
I disagree as do many, many others. There are assertions that the U.S. legal principle of fair use applies when training AI. This position has featured for some time in U.S. academic debate and in the courts. Perhaps it was this principle that prompted what court documents have recently revealed; namely that Meta employees, apparently with the approval of Czar Zuckerberg, had, with no payment to the books’ authors or publishers, downloaded approximately 82 Terabytes of pirated books from shadow (aka unlicensed) libraries to train AI systems
To pile Pelion on Ossa, Meta has announced that it will be using all public user data, including texts, photos and comments, to train its AI from June 2026, despite this practice having already been in play at Facebook and Instagram since 2007!
Fair use? You decide.
The fair dealing regime in the U.K. is more clear cut but, I suspect, no less abused. U.K. law provides a closed list of actions that do not require a licence. And, under E.U. law, text and data mining (TDM) exceptions to copyright protection are limited ( Article 3 of the 2019 instrument, Copyright and Related rights in the Digital Single Market – the DSM Directive).
TDM is ok when it comes to “……reproductions and extractions made by research organisations and cultural heritage institutions in order to carry out, for the purposes of scientific research, text and data mining of works or other subject matter to which they have lawful access.” So, not ok when Meta is building an AI powerhouse tool in the E.U., then.
Then there are the questions of whether authors and performers should be able to opt-in, or opt-out of their works being used to train AI; the latter a trickier prospect given the “that ship has sailed” problem.
And should such “opting” be the prerogative of the creators or their commercial partners?
And can the talent be sure any money that is generated (no pun intended) trickles down from corporation to creator? Phew.
The U.K. government has launched an AI consultation via the U.K. Intellectual Property Office. That effort, however, carried the stench of a fait accompli in the view of the talent and their representatives, it being tainted by the Government’s stated aims of a strong British AI industry.
To the fury of the creative industries, there were suggestions that the U.K. Government was going to introduce a new copyright and database exception which would allow TDM for any purpose… Fair? Fortunately, the prospect of an opt-out has now been put forward but, in the words of a respected British copyright academic “[w]hether it makes sense or not, is a different thing”…
Within the E.U., the approach, while not perfect, has tried to be more flexible and more creative.
For example, in responding to the E.U. DSM Directive, it was not long before the French and the German collective management organisations (CMOs) for music authors each took a stand. These included issuing proceedings against AI infringers, publishing intentions to opt-out and developing a licensing framework. The Irish music CMO, IMRO, went even further by introducing a painstaking exercise for their writer members right through their database. IMRO members can choose to opt-in.
Sadly, the large and influential U.K. music CMO has not followed suit. (PRS for Music seems to have something of a trust and confidence crisis on its hands, and may be distracted...)
Next question, then: How do we know what is and what is not an AI generated work?
A recent U.S. Copyright Office study published in January is helpful. Submissions to the Report largely, and in common with U.K. and E.U. legal principles, took the view that the current legal regime is adequate and that free and creative human choices remain a key requirement of copyrightability. The Report is usefully summarised in the Kluwer IP Law blog of April 24th, 2025. An AI-generated work based upon a copyright work can, in theory, itself be protected ONLY to the extent of the protected copyright work within it.
Incidentally, this reflects the position this author took in Ontario last autumn when speaking on just this subject at Canada’s annual International Institute of Communications Conference.
So, surely from a practical perspective, authors and performers (who have rights as well), as well as our wider culture, consent-and-reward should be part of the whole AI deal.
Within the Berne Convention Article 5 (2)’s wording is pretty unambiguous: “The enjoyment and exercise of copyright is not subject to formalities.” Copyright magically crystallises once an original work is expressed or fixed in material form. The initial owner of a copyright work is the creator or author; ownership may change but authorship remains the same. It is the owner that must be consulted and rewarded when a particular work is used. And therein lies a problem. In the words of George Merrill and Shannon Rubicam for Whitney Houston, “How Will I Know?”.
There already exists a global standard identifier for each category of copyright work. The ISO Office allocates identifiers that should apply world-wide. For music there is the ISWC for songs and ISRC for recordings. For books the ISBN sits on the spine of an analogue copy. An audio-visual work should bear an ISAN, though Hollywood, the BBC and others seem to be using their own and non-ISO labelling. Sigh.
Furthermore, and encouragingly, in May of last year 29 countries signed up to the ISO’s ISCC. The ISCC is an open-source technology that enables the identification, attribution and management of digital content across various platforms and industries. It provides a unique, decentralized and content-derived identifier for any type of digital media, including text, images, audio, and video. This is excellent news if, and only if, everybody uses it.
All well and good, but if there is no central registry of the authorship and ownership of global works…. back to Whitney. Is it now time to revisit this “no requirement for registration” within Berne?
An attempt to revisit this will be taking place in the U.K. on Wednesday, May 14th, 2025, at 1730 British Summer Time. This public event is being hosted by the Institute of Brand and Innovation Law within the prestigious Faculty of Laws at University College London.
The event will be accessible both in person in the UK and on-line.
This author must confess to being in the chair, but the panellists are stellar and cover a wide range of perspectives and nations.
For the U.S., UCL is pleased to welcome Jalyce Mangum, Attorney-Advisor at the Office of the General Counsel, U.S. Copyright Office.
Making sense of these registration conundrums from a practical perspective will be Krishna Sood, Assistant General Counsel, Microsoft.
The role of collective management of copyright works in this context will be addressed by Adriana Moscoso del Prado Hernández, General Manager, GESAC for the European Authors’ Societies.
And, to provide a thoughtful and distinguished academic perspective, the panel will be joined by Professor Martin Senftleben, Director of the Institute for Information Law (IViR), University of Amsterdam, in The Netherlands.
Fees for this event are scandalously affordable, available at between £8 and £35. Bookings can be made via this link below:
© Amanda Harcourt 2025
Honorary Professor of Practice, IBIL, Faculty of Laws, UCL
https://harcourt.global/ & at Blue Highway Advisory
Comments