The editor of the ‘New York Times’ warns of the hijacking of the public agora by AI

BarcelonaIn just over four years, artificial intelligence has managed to occupy a central place in our lives, but its current model threatens not only the sustainability of the media but also the foundations of democracy. This was argued on Monday night by A.G. Sulzberger, editor of the New York Times, during his intervention within the framework of the major annual meeting of Wan-Ifra, the main global press sector, which is being held this year in Marseille.

“The hijacking of public space is possible thanks to the original sin that animates AI products: a blatant theft of intellectual property that has occurred on an unprecedented scale,” warned the executive. “Tech giants plunder news websites without permission or compensation. They repackage this stolen property as their own, diverting audiences and revenue that would otherwise go to the journalistic organizations that created this content. And this doesn't happen just once during the training process, but countless times every day”.

Sulzberger denounced that while the companies behind the main generative AI tools are some of the richest companies in human history, they do not assume the responsibilities arising from their enormous impact and the control they exercise over citizens' data and attention. As he recalled in his diagnosis, applications like ChatGPT and similar ones need four ingredients to function: the talent of computer scientists, computing power, electricity, and data. To achieve the first three, these companies disburse large sums of money, while for the fourth, data and content, they take them without any fair compensation, which can end up eroding the system itself and harming them.

The fact that searches in AI chats provide information directly, without the need to go to the medium that generated it, and often without even citing it, means that the news outlet cannot monetize the investment of resources it has dedicated to producing that content by placing advertising on its website. And this happens once the big tech companies have managed to capture the bulk of brands' investment, with their ability to manage these ads in a personalized way, thanks to the vast personal data they handle. As the executive recalled, Meta alone – the company behind Facebook and Instagram – already bills eight times more than all the newspapers on the planet combined. And in two decades, media advertising revenue has fallen by 80%.

“I fear we are rapidly moving towards a future with fewer and fewer journalists willing to do the hard and difficult work of original journalism: going to places, talking to people, digging up information, covering important topics and events, providing context and analysis, and investigating the powerful,” lamented the editor. And he recalled that these effects are not limited to the press, but also affect all creative industries, which employ 50 million people worldwide and produce approximately $12 trillion in economic value.

In fact, when technology companies talk about “data,” as he admitted, they are already winning the battle of discourse, because in reality they are referring in most cases to “copyrighted content”.

Given this landscape, Sulzberger complained that the sector has bowed its head too much and has acted in a fragmented way against these companies. He advocated for mutual understanding: “Media companies should want to embrace what good AI can bring. And technology companies should support a healthy and sustainable stream of information and ideas and creativity that feeds AI.”

Negotiating power, however, is limited, as the six most prominent AI companies have a combined valuation of $11 trillion, which is three times the GDP of France. “This means that this intellectual property theft is not happening due to a lack of ability to pay for it. Although the agreements to license content that have been signed are not public, considering the small scale of those that have come to light, everything indicates that less than 0.5% of their investments will be dedicated to compensating the people and companies that create the data with which AI is run.”

Sulzberger recalled that five of the ten most used websites for training the most popular AIs are owned by press publishers. “Although most AI companies hide the sources they use to train their applications, the New York Times was the largest individual source of data with the owner of the databases used to train many different models.” The editor also revealed that the newspaper has already spent over $20 million on lawsuits over the last two and a half years, in particularly long and arduous proceedings. His intervention concluded with a call to act together, pressure lawmakers, claim the rights that belong to the media, and, in case of agreeing with these companies, to do so carefully so that short-term benefits do not come at the expense of compromising future viability.