UK licensing bodies have announced a “pioneering” collective licence that will allow authors to be paid for the use of their works to train generative AI models.
The Copyright Licensing Agency (CLA) – which is directed by the Publishers’ Licensing Services (PLS) and the Authors’ Licensing and Collecting Society (ALCS), representing publishers and authors – will develop the licence, set to be the first of its kind in the UK.
Expected to be made available to AI developers this summer, it will allow copyright holders “who are not in a position to negotiate direct licensing agreements with AI developers” to be paid for the use of their works.
“When we surveyed our members last year, they made it clear that they expect us to do something about their works being used to train AI,” said ALCS CEO Barbara Hayes. 81% of respondents said that they would want to be part of a collective licensing solution if ALCS was able to secure compensation for the use of writers’ works to train AI in cases where individual, case-by-case licensing is not a viable option.
The announcement comes as the UK government reviews responses to a consultation on its proposals for a copyright exemption for text and data mining, allowing AI companies to freely use copyrighted works unless rights holders opt out. The new licence “shows that a copyright exception is neither necessary nor desirable”, said the ALCS.
The government’s proposal “would give very limited choice, wouldn’t remunerate creators or provide any transparency about which works are being used”, said Hayes.
The collective licence, on the other hand, “will further demonstrate that licensing is the answer and can provide a market-based solution that is efficient and effective”, said Mat Pfleger, the CEO of CLA.
Training is transformative use.
It’s not like this ends with authors being paid every time a neural network gets used - we’d see a properly open model emerge. Smaller networks, trained on hand-pruned data, already look like the way forward. Generalization from sparse examples is how any of this works.
If people can’t simply let the robot read every book in the library, they’re not gonna pay for every book in the library. They’re gonna switch to some above-board torrent of public domain / Creative Commons / permissively-licensed content. That didn’t happen first because scaling up worked better, to a point. Aaaand scale prevents competition. Also it was a great excuse to spy on as much private data as possible.
When the bubble bursts, the tech will remain and flourish, and there’s not gonna be much money involved.
