NYT Claims Microsoft Built A Massive Supercomputer Just To Steal Their News

Following a recent Supreme Court precedent involving Cox Communications that drastically raised the bar for proving contributory copyright infringement, The New York Times moved to aggressively remodel its courtroom strategy against Microsoft and OpenAI. Plaintiffs now have to prove that an accessory deliberately incited infringement, which turned a boring cloud computing bill into an alleged multi-billion dollar conspiratorial setup.

Instead of treating Microsoft as a mundane cloud hosting provider selling server time, the newspaper’s amended complaint paints the tech giant as an active mastermind. According to NYT spokesperson Graham James, Microsoft did not just rent virtual space—it specifically architected a ridiculously bespoke supercomputer engineered to harvest the entire internet, with a disproportionate chunk of high-value journalism thrown into the digital blender.

The lawsuit claims this monstrous machine scraped protected archives with deliberate surgical intent, enabling OpenAI to construct modern language models that spit out full paragraphs behind paywalls without paying a dime. The paper points to millions of surrendered chat sessions where users casually bypassed subscriptions simply by prompting ChatGPT for the next paragraph, while Microsoft allegedly padded its market cap by a cool trillion off the resulting technology.

While Microsoft dismisses the pivot as a desperate legal scramble to dodge unfavorable precedents, OpenAI spokesperson Drew Pusateri insists that feeding the entire sum of human journalism into neural networks is pure, untouchable fair use. If the newspaper manages to convince a federal judge otherwise, the court might be forced to consider the nuclear option: ordering the total destruction of the foundational models.

The trillion-dollar dream of artificial general intelligence hinges on whether typing a query to extract proprietary news is transformative progress or just automated grand theft with extra computing steps.

Source: arstechnica.com