- Flurry of content licensing deals
- New potential revenue streams
Apple (US:AAPL) got into trouble this month because an advert for its new iPad showed musical instruments, statues, cameras and books being crushed by a hydraulic press. The video – which was supposed to spotlight the tablet’s ultra-thin design – unwittingly symbolised a big fear associated with Big Tech. Namely, that it is destroying human creativity in pursuit of profit.
The media sector has been vocal in its criticism of tech giants – particularly since the emergence of generative artificial intelligence. Nigel Newton, chief executive of Bloomsbury Publishing (BMY), said it was like the “Wild West out there”, with tech companies “pirating” copyrighted material to feed their algorithms. “It’s the greatest heist in recent history,” he told Investors’ Chronicle.
This is resulting in regulation, industrial action and a wave of lawsuits against the likes of Microsoft (US:MSFT) and OpenAI. Authors including John Grisham and George RR Martin, who wrote the Game of Thrones series, have already sued OpenAI, accusing it of “systemic theft on a mass scale”. The New York Times (NYT) is also suing OpenAI and Microsoft.
Some organisations are taking a different approach, however: if you can’t beat them, join them.
News Corp (US:NWSA), whose publications include The Times, The Sun and The Wall Street Journal, has just signed a multi-year content licensing agreement with OpenAI said to be worth $250mn over five years. The Financial Times, Axel Springer and Informa (INF) have made similar announcements in recent months, while social media platform Reddit (US:RDDT) has given OpenAI access to its user content.
Reach chief executive Jim Mullen warned about this type of deal in March. “The challenge we have as an industry is that we need to be unified,” he said. “If we stick together and work with [generative AI companies], then that’s a really strong position that we have – particularly with government – to help us get to there… It only takes one publisher to break away and start doing deals and then it sort of disintegrates.”
One also wonders whether the deals reflect fair value for what OpenAI will use: have they managed to embed royalties in licence agreements, for instance? Bearing in mind that the information could be used to train entire new products the publishers haven’t conceived of, that sort of provision would better future proof the business model.
Edison analyst Fiona Orford-WIlliams said there is a “huge danger – because rules have been ripped up – that companies will sell themselves short. Tech companies have deep pockets.” This worry has been echoed by Bloomsbury’s Nigel Newton, who wants the value of content licensing deals to increase.
From an investor’s perspective, uncertainty abounds. “It is difficult to formally evaluate the nature of the access that is being granted, and the restrictions on how content is being used, and the financial nuts and bolts of the deals,” said Shore Capital analyst Roddy Davidson. Information provided by publishers is simply too limited at the moment.
Regulators may have a tough time digesting all of this as big tech banks on its usual strategy of enmeshing technologies ahead of the rule-makers getting to grips with them. Firms and individuals protecting their interests can be relied on to be more dynamic, however.
High profile cases like that of Hollywood actor Scarlett Johansson, for example – whose voice has allegedly been copied and used for purposes she did not agree to – may spur regulators into action. At the same time, fears of fake pictures and videos made with its products have pushed OpenAI to build a tool to identify images made with AI to prevent abuse, ensure transparency on AI-generated content and improve access to accurate voting information.
Mining the archives
It’s not necessarily all doom and gloom, however. Parts of the media landscape may stand to benefit. FTSE 100 company Informa has signed a non-exclusive agreement with Microsoft, providing access to some of its academic texts in exchange for an initial fee of $10mn and a recurring fee across the next three years.
Such agreements “offer the potential for a nice windfall”, according to Thomas Singlehurst, head of European media equity research at Citi. Plus, there’s scope for bigger payments in the future. “The deal only covers a proportion of Informa’s back catalogue – maybe they can extend the relationship if successful. Or, because it’s non-exclusive, maybe they can sell access more broadly, ” he said.
Publishers with extensive back catalogues are worth keeping a particular eye on.
Despite its criticism of Silicon Valley, Bloomsbury said it “hasn't made any decision yet” about an AI partnership. Reports that Meta (US:META) considered buying publishing house Simon & Schuster last year to train AI models may be weighing on management’s mind.
Education giant Pearson (PSON) is another stock that might be tempted to form an alliance. Sales in its higher education division have been falling, partly because students are turning to “non-mainstream” publishers, and a licensing deal could provide a lifeline – not least because partnerships with tech firms are about more than data access.
With the help of third-parties, publishers have scope to transform their archives and create new learning tools. Informa, for instance, is collaborating with Microsoft to build a “citation engine” which will help students and academics generate references automatically.
Dangers ahead
Caution is the prevailing mood, however, which makes sense: by licensing content, companies may be compromising their ability to create their own AI tools or – even worse – cannibalising their front catalogue.
In its complaint, the NYT asks for the “destruction” of large language models (LLMs) such as ChatGPT that have used its content. Although, in Harvard Law Today, Mason Kortz of the Harvard Law School Cyberlaw Clinic, posited that the NYT is angling more for a settlement (like 98 per cent of US civil claims). With 16mn unique records of NYT content and statutory damages of up to $150,000 per work, that gives the publisher quite a big stick.
There are grey areas. For example, publishers use data under licence to produce charts and visualise trends. Data providers have yet to stipulate their policies regarding AI use-cases, either in terms of licensees utilising AI tools to manipulate and visualise data, or LLMs that make tertiary use of data they have gleaned from secondary sources like articles.
Realistically, the NYT action is unlikely to scupper LLMs, but its hard-nosed litigious approach may well ensure the newspaper eventually strikes one of the most lucrative deals for publishers. Reach boss Jim Mullen said in March his company was not in talks with OpenAI, but despite his call for unity, said it was a “very fluid process”. “We produce content which is really valuable and we would like to license or agree how the user base intelligence to actually inform their AI and their open markets,” he said.
The fractious relationship between tech and traditional media makes it hard to shake the fear that something will go wrong. In the early internet years, publishers lost huge chunks of advertising revenue to the likes of Google and Facebook. However, formalised agreements seem like a step in the right direction and – as noted by Citi’s Thomas Singlehurst – provide a “lovely validation that a good database trumps a good algorithm”.
- with James Norrington
