Deals with the Devil? Publishing’s Uneasy Alliance with AI 

AI is cutting a swathe through the media industries - but how is it impacting major publishers? John Still looks at the choice publishers face between licensing and legal wrangling, as they seek to take control of their content.

The rapid rise of generative AI over the last two years has left many industries facing existential questions, not least publishers. As AI seeks to get smarter, it needs data and information to consume and learn from, leading it to the doors of the internet’s publishers and major media organisations.

While some of the more obvious questions around creativity, productivity and elimination of jobs are a constant cause for discussion, a major dichotomy seems to be emerging: do publishers team up with the machines, or try to take them down?

The licensers

A number of major publishers have deals in place with AI companies such as OpenAI; they’ve seen the AI future and have decided to get something in return for their content before the horse bolts entirely anyway. As Steve Hasker, chief executive of Thomson Reuters, which has ‘a number’ of deals with AI companies puts it: “These models need to be fed. And they may as well be fed by the highest-quality, independent fact-based content.”

Dotdash Meredith, publisher of People, Investopedia, Food & Wine, and InStyle, recently signed a deal with OpenAI. In a letter to shareholders, IAC’s (Dotdash Meredith’s parent company) chief executive Joey Levin stated: “Some of the large incumbent technology companies may fight this reality for a bit, empowered in the near-term by dominant market share in other categories, but I expect most will eventually follow OpenAI’s lead, driven by either reason, reputation, litigation, or regulation.”

Dotdash Meredith is just the latest to sign a faustian pact with AI. Several major publishers have struck licensing deals with OpenAI to allow the use of their content to train OpenAI's AI models: Financial Times, Axel Springer (owner of Politico and Business Insider), Le Monde (a major French newspaper), Prisa Media (a large Spanish media conglomerate) and The Associated Press are some of the big names to have jumped on the AI bandwagon, with the deals reportedly ranging from USD$1m-USD$5m (£787,000-£3.9m) per year for the publishers.

For the AI companies, the goal is to be legitimised through their association with major publishers, and bolstered by claims to be trained on ‘quality’ data. With decades of content output, licensing deals allow AI companies to boast a treasure trove of information, that their generative AI is trained on the best of the best.

In some ways, the OpenAI/DotDash Meredith deal sets the bar for the licensing argument. The terms, a one-off payment for legacy content plus an ongoing ‘subscription’ to new content answered the question of how to price up decades of publishing and still maintain revenue going forward, a way of future-proofing the deal rather than a simple lump sum sale of their IP.

But aside from future proofing, what’s in it for the publishers? AI-driven analytics and predictive modelling enable publishers to gain deeper insights into audience behaviour, content performance, and market trends. By analysing vast troves of data in real-time, publishers can identify emerging topics, optimise content distribution strategies, and anticipate reader preferences. This data-driven approach empowers publishers to adapt swiftly to changing market dynamics, optimise resource allocation, and capitalise on new opportunities for growth.

Another prize being proffered is in ad tech, and advancing publishers own first-party data and audience targeting propositions. The Dotdash Meredith deal included attributed links being featured on ChatGPT, which reaches 100 million users, but having OpenAI’s capabilities augmenting the publisher’s D/Cipher intent-based ad targeting platform is the golden goose. Across the programmatic supply chain, demand from buyers for high-quality targeting driven by AI is red-hot, with buyers citing this as the main driver towards using AI tools. Get paid by AI firms, and reap the rewards of higher demand from marketers. Win-win?

As AI innovation spreads in publishing, content licensing deals that serve to supercharge ad offerings will become more popular. Advertising tools like Wall Street Journal’s Thematic AI aimed at increasing operational efficiency would benefit from the boost offered by the AI giants.

The litigators

While many publishers are playing ball, some are trying to stop the unlicensed use of their intellectual property through the courts. 

The highest profile of these so far is the New York Times - the publisher filed a lawsuit in December 2023 against OpenAI and Microsoft for copyright infringement, over the companies’ use of its content to train generative AI and large-language model systems. The New York Times claims that its content was given “particular emphasis”, and that OpenAI sought “to free-ride” on the publication’s large investment into its journalism by building products without permission or payment. The lawsuit calls on the companies to destroy any chatbot models and training data which use copyrighted material from the Times.

The New York Times are being followed in their legal quest by eight newspapers owned by Alden Global Capital, including New York Daily News, Chicago Tribune and others. Frank Pine of MediaNews Group, an Alden subsidiary, didn’t pull any punches: “The misappropriation of news content by OpenAI and Microsoft undermines the business model for news. These companies are building AI products clearly intended to supplant news publishers by repurposing purloined content and delivering it to their users.

“Even worse, when they’re not delivering the actual verbatim reporting of our hard-working journalists, they misattribute bogus information to our news publications, damaging our credibility.”

A fine line

Even with questions about employment opportunities and the future of journalism, the choices between licensing and litigating lays bare the complex interplay between tech innovation and intellectual property rights. 

As AI systems increasingly rely on vast datasets, often sourced from published works, the necessity for fair and enforceable licensing agreements has become clear. The AI current is strong, and no one can blame publishers for looking to receive a return before that tidal wave has already crashed.

But equally, publishers’ works are worth defending. The legal route is a crucial check on AI’s progress, compelling AI companies to engage in more transparent and ethical practices, and perhaps pushing the industry toward more standardised and mutually beneficial licensing agreements going forward.