Executives at The Guardian, Financial Times and Le Monde agree publishers must get acknowledgement – and payment – for their news content.
Otherwise, Le Monde chief executive Louis Dreyfus warned at the Deloitte and Enders Media and Telecoms Conference 2023 last week, it could spell “the end for our business model”.
The news bosses all expressed fears over the use of the publications’ content in training AI large language models like ChatGPT without any licence or payment model in place – let alone credit. Some, including Telegraph Media Group chief executive Nick Hugh, also warned about the risk to trust from generative AI content.
Speaking on a panel about news sustainability, but dominated by discussion of artificial intelligence, Le Monde’s Dreyfus described the situation as “an emergency” and said “we need to define a common position” and establish “what AI companies are ready to put on the table for use of all content”.
Related
“Otherwise it’s the end of our business model. If somebody can type a question, or write stories, using our content or mixing it with some low-quality content, it’s a risk for the political debate, political society.”
Asked what opportunities he saw for journalism in the age of generative AI, which can create lengthy texts from short prompts by predicting which word is most likely to follow in a sequence, the French newspaper boss said: “I strongly believe there is more risk than opportunities for us. Our business model is based on trust. With AI, we are taking the risk of less reliability and less exclusivity.”
He said there was a risk to trust if a large language model was trained on both high-quality information sources and less authoritative ones.
AI and news publishers: ‘We’ve got to get control of our IP and manage it fast’
Jon Slade, chief commercial officer at the FT, said “there’s very good evidence” that his paper’s archive had been used to train large language models.
“And it’s been used without our consent, it’s been used without a licence. It leaves me slightly conflicted – but only slightly conflicted – because as a chief commercial officer I think there should be a payment or whatever for that.
“But on the other hand, I’m glad that our quality journalism is part of these models. And that may be one way to get down the risk of misinformation, but it doesn’t detract from the issue of the licence where it’s necessary.”
Slade likened the lack of payment to publishers to the period in which news publishers began to widely publish online free of charge.
“The key issue for publishers, 20 years ago, was that a lot of them gave their content away for free. And they can do that again. And we would see a similar level of disintermediation, or probably a much greater level of disintermediation, through [our content’s] synthesising and use in products that we don’t control but we have contributed to.
“So I think we have to remember that we’ve seen the movie before. We’ve got to get control of our IP and manage it fast.”
He said regulation would be “slightly useful in that sense”, but that he was worried “that the tone of some of the regulatory conversation seems to index more, currently, towards ensuring that AI businesses have access to data, rather than protecting existing copyright and IP regulation that already exists”.
Guardian Media Group chief executive Anna Bateson similarly said she would like to see both payment and attribution for her publications’ content from AI companies.
“If our content – that it costs a considerable amount of money and investment to produce – is being used to train models which are then incredibly powerful, and the basis for incredible value creation, then there needs to be some sort of acknowledgement of that.”
[Read more: Journalists – ChatGPT is coming for your jobs (but not in the way you might think)]
Of trust, The Telegraph’s Nick Hugh said: “If search will evolve into AI that would have a fundamental impact on the delivery and the access of journalism… How do you know what’s true and what’s not true?
“There’s not enough clarity and it’s therefore creating an opportunity for bad actors. That, to me, is the big risk that you’ve got. But obviously, there are opportunities. Although I’m much less pro-automated content generation.”
How publishers are incorporating AI into their work
Several of the publishers said their teams are using non-generative AI in parts of their workflow but not using it to write editorial content.
For example, Dreyfus said that at Le Monde’s recently-launched English language edition an AI carries out the initial translation before a human translator takes a second pass and then an English-speaking editor does the final check.
He added that “we need a process and transparency and rules which are known to a reader” for involving AI in any writing.
Slade, meanwhile, ruled out the use of generative AI in creating FT content.
“We’re very clear that we’re not going to have articles written by AI,” he said. “Articles in the FT will be written by journalists.
“Defining what writing means, and defining that process actually, needs a little bit of work, though. Because we’re sourcing information using AI. We’re going through huge troves of data and seeing where the stories and where the patterns are…
“But the final words, the final editing, that’s always going to be done by a human, we’re clear about that.”
Nonetheless, the company is spending more on AI this year than it was last year: “We need to get our governance right first, but yes, we’ll certainly be investing in it. And we’re investing mainly around the area of discovery.”
The Guardian, Bateson said, has established a working group tasked with determining the principles by which the paper will use the technology, and that it had been conducting training, “particularly with our journalists, about how to think about it and what we think is okay from an experimental point of view [and] what isn’t”.
[Read more: FT creates AI editor role to lead coverage on new tech]
Shifting revenue streams
Away from the subject of AI, the executives also discussed how their revenue and subscriber patterns have evolved in the past year.
Bateson said The Guardian now generates 70% of its revenue from digital and 35% of it from outside the UK. A quarter of reader revenue now comes from the US, its fastest-growing market.
Slade said the US is also the fastest-growing for the FT, having seen a 16% rise in its paying readership there last year. The UK was only narrowly behind with a 15% increase.
Slade added that the financial publication is drawing 90% of its revenue from digital sources and that its “total global paying audience” – including from the newspaper, FT.com, FT Chinese, FT Specialist and paying delegates for FT Live – stands at 2.4 million. He noted that the figure is more than four times higher than the FT’s all-time highest print circulation of 450,000.
Dreyfus meanwhile said that in the previous three years, Le Monde has doubled its digital subscription revenue to €16m – but that the paper needs a strategy for reaching a younger audience: “That’s what keeps me up at night.”
And The Telegraph’s Hugh described himself as “hugely bullish on the outlook, actually, for quality news as a whole”.
Hugh said The Telegraph’s subscription volumes are “just a shade under a million” and that he expects to cross the million mark before the end of the year. “Four or five years ago that was less than 300,000, so that’s a great number for us.”
He said the paper’s earnings before interest, taxation, debt and amortisation (EBITDA) is “in the range” of £60m, which he said, “as a percentage is a very good margin”.
He commented that what worries him – though “it’s more of a positive story”, he said – was that “there’s quite a lot of opportunities for us. And how do we make sure that we really focus on a handful of opportunities rather than trying everything”.
A resulting focus on growing subscriptions had, he said, been “the genesis” of The Telegraph’s acquisition of Chelsea Magazine Company.
Email pged@pressgazette.co.uk to point out mistakes, provide story tips or send in a letter for publication on our "Letters Page" blog
