Axel Springer’s OpenAI deal sets new template for media ties with Big Tech

The landmark content licensing deal between German publisher Axel Springer and Microsoft-backed OpenAI marks a breakthrough in the media industry’s efforts to secure its commercial future as artificial intelligence technology takes its next generational leap. 

Media groups not only face an existential threat from generative AI, which creates text, images, audio and video that is indistinguishable from that produced by humans, they are also fighting over Big Tech’s use of their content to train AI software.

Executives have focused for the past few months on ensuring that, unlike in the early years of the internet, Big Tech fairly compensates the media industry.

This week’s deal is being scrutinised by rival publishers as a potential road map for how the relationship might work in the future.

The agreement, worth tens of millions of euros a year, will allow the ChatGPT creator to use content from titles such as Politico, Bild and Business Insider. It reflects much of what the industry has been puzzling over all year: how to value content archives that can stretch back for decades while creating an income stream from new journalism. 

Media analyst Ian Whittaker described the deal as “a model for everyone else — flat fee for the historic data plus ongoing annual fee”. 

Over the past year, large AI companies including OpenAI, Google, Microsoft and Adobe have met news executives from companies including News Corp, Axel Springer, The New York Times, Guardian Media Group and the Financial Times to discuss copyright and licensing issues around AI products such as text chatbots and image generators.

GMG said it hoped the Axel Springer deal “marks a change in direction whereby [AI] model developers that use copyright content to build and inform their products enter into commercial agreements with copyright owners that invest in the high-quality media, news and information”.

The early years of the internet represented a huge shift of value from the media to the technology industry as companies such as Google and Facebook used freely available content to help build multibillion-dollar advertising empires. 

People familiar with the discussions between Axel Springer and OpenAI directly linked the deal with the media group’s intentions to prevent a similar outcome as rapidly proliferating AI-based chatbots siphon traffic away from news websites.

Axel Springer boss Mathias Döpfner has had longstanding tensions with the tech industry, calling for closer regulation of industry giants such as Google over the risks to individual rights and competition. In a 2021 letter to European Commission president Ursula von der Leyen, he argued that “the absolute dominance of tech platforms also means the disappearance of diversity in journalistic, artistic and commercial products and services”.

Other news groups are expected to swiftly follow Axel Springer’s lead. News Corp chief executive Robert Thomson said last month that the owner of The Times, The Wall Street Journal and The Sun newspapers was “looking to the future in maximising the value of our premium content for AI”. 

The group was “in advanced discussions with a range of digital companies that we anticipate will bring significant revenue in return for the use of our unmatched content sets”, he said, referring to a “new phase of negotiations with the rise of generative AI”.

Others remain more cautious. “Clearly it’s for every media company to decide what suits their strategy,” said one senior industry executive who asked not to be identified. “That will depend on the extent to which they are relying on reader or advertising revenue.”

Axel Springer has struck a two-tier payment scheme in which it will receive a one-off fee for its historical content used for AI training and a recurring income stream from the use of new content. 

There are also “kickers” — in effect extra payments — for popular content, meaning the media group will be paid more each time its articles are used by AI. Other models — such as using word count to determine compensation — have been discussed by the industry in the past. 

The deal is not exclusive, which means Axel Springer can strike similar agreements with other generative AI companies.

Analysts say it is difficult to judge the deal commercially given the financial terms were not disclosed. A person close to the talks said Axel Springer would be getting an “eight-figure sum”. 

One media executive also pointed out that, regardless of the particulars of this deal, “monetisation happens best when it happens on your website”. If the consumption of quality journalism is happening on someone else’s platform long term, “that doesn’t strike me as being a good thing”. 

The executive added they were “surprised to see OpenAI had struck this deal because that’s not the tone that Microsoft has been striking”, noting that AI companies had previously shown they did not expect to have to pay for media content, “because [they] don’t think [they] are infringing copyright”.

But a deal makes sense for OpenAI, not just in giving it access to a trove of information that can train its future models but also for any future regulatory issues the sector may face. 

AI development requires large libraries of written works to “train” or improve systems. Better-quality information such as paid-for, premium journalism should in theory be worth more money.

Douglas McCabe, an analyst at Enders, said that “journalism as an AI input is highly attractive . . . the training runs wide and deep: not just a bunch of new facts and commentary, but cultural references, the way language communicates with nuance, through tone and phraseology, a hierarchy of cultural ideas and interrelated themes”.

The Axel Springer deal, he said, quantified this principle “in a way other media companies will welcome”.

OpenAI this year agreed a deal to license Associated Press’s archive of news stories going back to 1985 to help train its models but it has otherwise used freely available internet content for ChatGPT. 

Axel Springer said it had previously, like several other publishers, blocked OpenAI’s software from trawling its websites for training material. However, industry insiders say it is “next to impossible” to gauge the extent to which journalism has already been ingested for the training of AI models. 

This has led to the threat of both legal action — such as from groups of authors including comedian Sarah Silverman — and regulatory oversight. The US Federal Trade Commission has opened an investigation that will in part look at whether OpenAI engaged in unfair or deceptive privacy or data security practices in scraping public data.

“AI developers once seemed to have the upper hand but the balance of power is shifting in favour of content creators,” said Gill Dennis, an intellectual property lawyer at Pinsent Masons in London. “Without access to these works, AI development cannot go ahead.”

This tacit acknowledgment of value for journalism in the next stage of technological development is as welcome as the particulars of the financial settlement, according to media executives who hope this week’s deal will set the path for other large AI providers to follow.

“However artful the artificial intelligence,” said Thomson last month, “it is no match for great reporting and for genuine journalistic nous”.