The most vocal critics of generative AI view the technology as overt labor theft, due to how the underlying web scraping and AI training works, and internal documents from Microsoft and OpenAI revealed by The New York Times v. Microsoft and OpenAI echoed these concerns.

One internal Microsoft document described it as "an astonishing theft of unprecedented proportions," while ChatGPT development head Nick Turley posted in a June 2023 memo that AI posed "an existential threat" to publishers since AI becomes "more and more substitutive" as it gets better. Another Turley memo from that year in February stated that "no matter how prominently we show the (source) links, users won't click" and that AI products are "largely substitutive, period," a statement going against the fair use doctrine of transforming but not replacing the original work.

Additionally, Microsoft employees internally debated whether or not OpenAI was committing "the largest theft of labor in human history" that could also create a "doom loop," since AI would be left unable to grow after cannibalizing all of its training sources out of the market. Internal memos written by Brent Hecht, Microsoft's Director of Applied Science, warned that large AI models "are a product that destroys its supply chain."

The New York Times' office entrance in Manhattan, New York. Image: Alberto-g-rovi on WikiMedia Commons (CC 3.0 License)

The New York Times may have just found the smoking gun it and the other publishers suing Microsoft and OpenAI need to win in court, and we have seen other victories against large AI players before.

While Microsoft's official position is that Hecht was hired specifically to "present divergent and asymmetric perspectives," it's hard not to see where he's coming from. While OpenAI and Microsoft argue that their AI remains fair use since it transforms existing works into "entirely new work" that don't harm the value of the original work, there's a lot of contention there.

Not only have several AI trends infamously ripped off the styles of prominent anti-AI artists, like Hayao Miyazaki, but it's hard to see the fair use argument when people at both companies were clearly aware that what they were doing could be harmful to the health of the open web and publishing as a whole. It has proven harmful to web publishing revenues across the board, impacting outlets large and small. This isn't to mention the environmental or safety problems posed by AI, or Amazon's practice of scanning and destroying rare books exposed by 404 Media.