Freelance journalists do not agree with their work being used to train AI, and most would like to be compensated for work being used in this way, a National Union of Journalists survey has revealed.
The survey comes as many publishers have signed deals with AI companies that allow their content to be used as reference points for queries through chatbots or for the initial training of such chatbots.
Freelance journalists are not typically compensated in such deals, with contracts often including opt-outs around the use of content in “technologies”.
Even in the rare deals that do compensate journalists for the use of their work, such as at Scandinavian publisher Schibsted, the payments are limited to staff only. Le Monde in France is sharing 25% of AI licensing revenue with staff.
NUJ general secretary Laura Davison said “urgent change” is needed around both consent and compensation when it comes to the use of work by AI.
More than half (60%) of freelance journalists believe their work should be licensed for uses such as AI training and AI inference (where the AI model answers a specific query with reference to an article) only with their explicit consent, a survey of 414 journalists by the NUJ and the Authors’ Licensing and Collecting Society discovered.
Just 11% of the remaining journalists surveyed said they agree their online work should be licensed for use by AI (with or without their consent).
A fifth of freelance journalists (20%) said their work should not be licensed for use by AI under any circumstances.
Journalists are also keen to be compensated for past use of their work for training AI, the survey found.
Journalists were slightly more likely to hope to be paid via collective licensing (59%) than by taking action for copyright infringement (50%), with just 5% saying they would not seek compensation for past use.
Journalists are increasingly using services such as “Have I Been Trained?” to see if their work has been included in data sets used to train generative AI models to produce text, images and other media.
That site is currently down, although The Atlantic offers a tool for writers to search if their books or scientific papers were part of the pirated Libgen data set on which Meta trained its AI systems.
Based on responses from print, video and photo journalists, 29% had searched to see if their work had been included in such training datasets.
Just over a fifth of journalists (21%) found evidence that their work had been used to train AI.
‘Opt-in mechanisms for journalists needed’
Davison said: “For too long, AI developers have used copyright journalistic material without consent to train their models.
“We need urgent change, with clear opt-in mechanisms allowing journalists to make informed decisions about how their content will be used in AI technologies.
“The unfair imbalance of well-resourced tech companies against freelance journalists cannot continue to go unchecked. We continue to urge fair remuneration and greater transparency over the sources of training data.”
Even without specific deals struck by publishers, freely available (ie not paywalled) information on the internet has likely been used to train large language models such as ChatGPT.
OpenAI says, for instance, that its foundation models are developed using “information that is publicly available on the internet”, alongside information that it partners with third parties to access.
The organisation says: “For publicly available internet content, we use only information that is freely and openly accessible on the internet. We do not intentionally gather data from sources known to be behind paywalls or from the dark web.”
OpenAI did not respond to a request for comment.
Richard Combes, deputy chief executive of ALCS, said: “In the age of AI, freelance journalists face an uphill struggle to ensure their rights are recognised, protected and remunerated.”
Last year, ALCS launched SCOOP, an organisation campaigning for freelance rights in the AI era, along with the non-profit licensing agencies representing photojournalists – DACS and PICSEL.
Combes said: “In this evolving landscape, the focus for SCOOP will be ensuring that freelancers collectively have the option to secure payment for works that have already been used by AI companies, either through unlawful online scraping or under licences entered into by publishers.”
SCOOP aims to “reset the imbalance” by ensuring that freelance journalists in the UK receive their “due share” of revenues generated from the secondary use of their works.
Current situation follows years of freelance journalists ‘signing away’ their rights
Freelance journalist Claire Coleman, who has written for the Daily Mail, Daily Telegraph, The Times, Metro and others for 20 years, wrote to her local MP Bell Ribeiro-Addy to voice her concerns over the use of copy to train AI, and in particular the Government’s plan to allow AI companies to use copyrighted materials unless rights holders ‘opt out’.
Ribeiro-Addy responded: “I understand the concerns surrounding copyright infringement, specifically the use of copyrighted material by AI developers to train their systems, often through processes like text and data mining.
“It is crucial that the voices of creatives remain at the centre of these advancements. Creators, as the rightful holders of their content, should have a leading role in shaping the framework that governs how their work is used, ensuring that innovation happens in a fair and just way.
“While I acknowledge the Government’s decision not to extend the text and data mining exception to cover the development of AI for commercial purposes, I believe that the exception still plays a vital role in enabling researchers to make copies of copyrighted material for the purpose of computational analysis – provided they have the right to access the work. This is an important distinction to maintain as it strikes a balance between fostering innovation and protecting the rights of content creators.”
Coleman told Press Gazette she first became aware of issues around copyright through friends who are authors, which “triggered an understanding of the way in which people’s words were being used without their consent”.
She said she feels that the current situation has arrived after many years in which journalists became used to “signing away” rights rather than getting syndication payments, as used to be the norm.
Coleman said: “I just think now, as a freelance journalist, you get used to signing away your rights in order to be able to get paid, and I think there are certain battles that have been ongoing on that front for a while. I feel like the first one came when suddenly it was ‘Oh, your stuff’s not just going to be published in a newspaper. It’s going to be published online as well.’
“Sometimes freelancers have the energy to battle those sorts of things, and sometimes we don’t. I think the reason that the AI training aspect of things feels like it hurts more is because it feels like any freelancer at the moment, I’m finding it harder and harder to get work.
“There seems to be less work being commissioned externally, and what is being commissioned is not being paid as well as it was previously. And so it feels like we’re fighting an awful lot.”
Claire said she believes AI may reach a point where it can replicate – or even outperform – the work of creatives whose efforts it has been trained on.
Coleman said: “I think AI will make it harder to get work. I think we’re already seeing it happening. I feel like if publishers are doing deals with these companies, then we have shot ourselves in the foot already, because I presume all of the contracts that we’ve already signed means that they can make money from work that we’ve already handed over to them.
“I appreciate that ALCS exists because it feels like the individuals are increasingly impotent, and have been for a really long time.
“There’s been so much talk of all the things that journalists grumble about, payment on publication, for example. Nothing ever happens. It doesn’t change because there aren’t enough of us to hold the line.
“Laws already exist. They’re just not being enforced. I’d like to see the Government not prioritise technological developments over existing businesses, and see writers recompensed in the same way they would be if their work was being used anywhere else.”
Email pged@pressgazette.co.uk to point out mistakes, provide story tips or send in a letter for publication on our “Letters Page” blog
Source: Freelance journalists want control over AI using their work, survey reveals
