OpenAI and Microsoft Proceed Despite Knowing It's a 'Devouring Cycle'

Overview

Recent internal documents unsealed in the lawsuit against OpenAI and Microsoft, brought by The New York Times, have shed light on the extent of big tech companies' awareness as they pushed forward with generative AI. The files reveal that OpenAI and Microsoft not only knew their data scraping constituted the "greatest labor theft in history" against content creators, but also acknowledged that it was initiating a "destructive cycle" undermining the foundations of the online economy. Despite privately recognizing these risks, company executives persisted with the strategy, prioritizing commercial interests over the sustainability of original content.

Shocking Revelations from Internal Documents

The tone of Brent Hecht, Microsoft's Director of Applied Science, was particularly stark within the 92-page disclosed documents. He described the data collection practices of ChatGPT and Copilot as the "greatest labor theft in history" and accused Microsoft's "fair use" defense of being a "complete mockery" of the legal concept. Although Microsoft attempted to distance itself, calling this an individual employee's personal view, another document showed that Microsoft Data Strategy Head Jordan Universal positioned Hecht as an employee holding "adversarial views," which instead corroborated that these concerns were not isolated within the company.

More crucially, the documents revealed senior leadership at Microsoft, including CEO Satya Nadella, had candid recognition of the "destructive cycle." Internal documents explicitly stated that the AI content strategy was forming a self-destructive loop: "Our products threaten the economic foundation of their basic suppliers; this is the situation we have created for the LLM business and its 'content supply chain.'" This meant that AI models are gradually destroying the source of content they rely on by replacing human creation.

Systemic Impact on Publishing

OpenAI was equally aware of the disruptive impact of its technology on traditional journalism. Jack Clark, OpenAI's Director of Policy, had warned that the company was building systems capable of substituting "the labor of people who define society's culture," likening ChatGPT to a "modern newsstand." Former OpenAI employee Nick Trli pointed out that once users obtain answers through chatbots, "there is no reason to click back through links to the source," directly causing a significant drop in referral traffic to media websites.

Furthermore, Cofounder Greg Brockman's pursuit of "trillions of dollars" in commercial interests stood in sharp contrast to employees' concerns over copyright infringement. Internal testing acknowledged that GPT-4 was "very good at regurgitation," able to copy protected content from outlets such as The New York Times word for word. Although the company claimed it would prevent memorization, its training data still contained a large volume of unauthorized and uncompensated material, including content scraped from behind paywalls. These pieces of evidence indicate that despite knowing they would irreversibly harm the publishing industry, big tech companies chose to accelerate the process.