If you can’t beat them, join them. That’s what content providers including news publications are currently doing as they battle to protect their business models from big tech’s artificial intelligence (AI) exploits, such as OpenAI.
Today, the Financial Times (FT) sealed a licensing deal with OpenAI, the developer behind ChatGPT, which will allow users of the AI chatbot to access summaries, quotes and links to FT articles.
Although the value and terms of the deal are confidential, OpenAI is reportedly offering news organisations as much as $5m per year to license their copyrighted content to train its AI models.
“This is an important agreement in a number of respects,” said FT group chief executive John Ridding. “It recognises the value of our award-winning journalism and will give us early insights into how content is surfaced through AI.
“We have long been a leader in news media innovation, pioneering the subscription model and engagement technologies, and this partnership will help keep us at the forefront of developments in how people access and use information,” he added.
The FT is not the first to strike an agreement of this kind, however.
In July last year, Associated Press was the first to sign a deal with OpenAI, which licensed its content archive that goes back to 1985 to the Microsoft-backed tech company for training. The companies are also looking at “potential use cases for generative AI in news products and services”.
Then, in December, Politico and Business Insider owner Axel Springer also agreed on a partnership with OpenAI to allow ChatGPT to summarise all its content, so long as it provided links and attribution.
A couple of months ago, the boss of Reuters said the company had struck “a number” of deals with AI companies looking to use its news content to train their models, although he remained tight-lipped on details.
The boss of News Corp, which owns The Times, has said it is looking to do some deal making of its own with AI companies.
And, while not a traditional news publisher, Reddit has struck a deal allowing Google to use its media content for AI tool training, reportedly valued at around $60m a year, according to Reuters.
How did we get here? A few decades ago, as news publishers began transitioning their content online, some initially offered it for free. But the advent of Google search in the late 1990s highlighted the challenges of sustaining a free model, prompting many publishers to adopt a paywall system.
As media analysts at Enders have put it in a new report: “Even if publishers can opt out of their content being used for training new models, they cannot currently opt out of contributing to generative search responses without also opting out of being included in search results altogether.”
Although the dominance of Google search still looms over publishers, the newest threat from Silicon Valley is generative artificial intelligence (AI).
ChatGPT and Google’s Bard have trained their generative AI models by scraping data indiscriminately from across the web.
Content creators are angry that their intellectual property has been and could be used in future without any consent or compensation.
“The first thing I would say is they need to start paying us,” The Times editor-in-chief said last week, Press Gazette reported. “They’re ripping off our content, and we would like them to pay for it. And once we get that out of the way, we’ll be even more excited about it,” he added.
It has led publishers to enter negotiations with tech companies to try to reach some mutually beneficial agreement.
As in the cases above, this has worked. But in December, the New York Times filed legal action against OpenAI and its parent company, Microsoft, after licensing talks broke down. The NYT is alleging copyright infringement and is seeking billions of dollars in damages.
The lawsuit has accused the tech giants of exploiting its “massive investment in its journalism” and threatening its ability to deliver “trustworthy information, news analysis, and commentary.”
British laws around copyright and fair use are unclear as they were created before AI was around. Government plans for guidance on copyright and AI initially failed and have since stalled.
Enders analysts continued: “Publishers collectively hold many cards at this moment, which should not be given away in a hurry despite their exact value being difficult to estimate.
“At a time where the financial benefits of licensing deals are equivalent to a small revenue boost to most, careful thought should be given to terms that fit in with the prevailing commercial goal of sustaining quality journalism and content creation online.”