An Unsettling Future for both Media and Tech: Financial Times' Partnership with OpenAI

Another week, another tense agreement between Big Tech and the media that report on its participants. In order to use the Financial Times' journalistic output as training data for its huge language models—which in turn power chatbots like ChatGPT—OpenAI announced on Monday that it has reached a licensing agreement with the newspaper.

After reaching agreements with publishers in the U.K., U.S., France, Spain, Germany, and other nations, OpenAI has now reached an agreement with The FT. These license agreements provide newsrooms some hope during what has been an exceptionally difficult period for the media industry.

"This agreement is positive for the Financial Times and could serve as a model for other major news organizations seeking to negotiate for small-scale extra income," comments Rasmus Kleis Nielsen, the head of the University of Oxford's Reuters Institute for the Study of Journalism.

Though the specifics of the agreement are still unknown, media outlets should be encouraged that OpenAI is willing to pay for this deal. OpenAI's most recent agreement also coincides with a shortage of high-quality training data for tech companies, as most of the internet has already been taken over by LLM providers, who have locked out the rest with "do not crawl" requests.

A study led by the late University of Cambridge researcher Ross Anderson warned that in the absence of high-quality human-generated training data, LLMs could experience "model collapse," wherein they are trained on their own outputs and begin to malfunction, much like inbreeding affects mammals. Previous research suggests that AI companies could run out of training data by 2026.

Another week, another tense agreement between Big Tech and the media that report on its participants. In order to use the Financial Times' journalistic output as training data for its huge language models—which in turn power chatbots like ChatGPT—OpenAI announced on Monday that it has reached a licensing agreement with the newspaper.

After reaching agreements with publishers in the U.K., U.S., France, Spain, Germany, and other nations, OpenAI has now reached an agreement with The FT. These license agreements provide newsrooms some hope during what has been an exceptionally difficult period for the media industry.

"This agreement is positive for the Financial Times and could serve as a model for other major news organizations seeking to negotiate for small-scale extra income," comments Rasmus Kleis Nielsen, the head of the University of Oxford's Reuters Institute for the Study of Journalism.

Although the specifics of the agreement are still unknown, media outlets should be encouraged that OpenAI is willing to pay for this deal. Additionally, OpenAI's most recent acquisition comes at a time when tech companies are struggling to find quality training data because LLM providers have already taken over most of the internet and are keeping the rest locked out due to "do not crawl" requests.

A study led by the late University of Cambridge researcher Ross Anderson warned that in the absence of high-quality human-generated training data, LLMs could experience "model collapse," wherein they are trained on their own outputs and begin to malfunction, much like inbreeding affects mammals. Previous research suggests that AI companies could run out of training data by 2026.

The FT report is released at a moment when AI firms are juggling a plethora of legal concerns about their training data. Eight newspapers that are part of the hedge fund Alden Global Capital were sued by Microsoft and OpenAI just this week for allegedly utilizing their material as training data. Independent researcher and consultant Lukasz Olejnik adds, "I see this as an approach of tech companies to alleviate some of the copyright infringement they risk."

Olejnik is curious, though, if any remarks he makes to FT reporters—he has been a source there for over ten years—will alter ChatGPT's impression of him and the questions it poses regarding him.  In fact, a data rights advocacy organization called Noyb filed a data protection complaint in Austria last week, voicing worries that ChatGPT inquiries may provide inaccurate information about specific persons.

One noteworthy item that the Financial Times' own report on its historic purchase omits? a monetary worth.

The absence of financial disclosure on the license agreements makes it challenging to determine the market worth of the back catalogs of publications. The Associated Press was among the first to sign a contract with ChatGPT's creators in July 2023, though they withheld the financial details at the time. According to reports from The Information, as of January 2024, OpenAI was reportedly giving publishers yearly deals between $1 million and $5 million, which equates to pennies per word for their work. (When Fast Company pressed the FT for the specifics of the contract, the FT refused.)

You're not alone if you think that's low. The New York Times is suing OpenAI for damages, alleging that the company violated copyright laws by using its journalistic content to train its tools without authorization. The Times claims that OpenAI is liable for "billions of dollars in statutory and actual damages," however it hasn't said how much money would be needed to settle the case.

The FT is the most recent company to agree to a settlement with OpenAI, while The Times is still pursuing legal action to get more money. However, there's a broader concern in the media sector that when well-known publishers sign agreements with startups like OpenAI, there won't be much money or interest left over for smaller publications.

Companies are now pursuing arrangements on their own, which is different from discussions held about a year ago when major publishers were purportedly attempting to form a coalition in order to obtain a more powerful negotiation position with OpenAI. Fast Company is aware that opinions have shifted due to worries that these alliances can be seen as price-fixing by anti-competition authorities.

According to Nic Newman, Senior Research Associate at the Reuters Institute, "Big platforms are doing limited deals with a few of the biggest, most trusted news sources in a few big companies, including news agencies. The disparities already present in the larger ecosystem are probably going to get worse as a result of these deals."