The New York Times has amended its lawsuit against OpenAI, directly targeting Microsoft for creating a specialized supercomputer allegedly designed to train AI models on copyrighted content without permission. Originally filed in 2023 as the first major publisher action against OpenAI, the complaint now argues Microsoft constructed an "unusually complex" machine specifically to mine the internet—with disproportionate focus on Times journalism—for building ChatGPT.

According to the updated filing, Microsoft's custom-built system wasn't generic cloud computing but a targeted tool to "train the most capable LLM in history" using high-quality journalism. The NYT claims this architecture allowed OpenAI to weight its articles more heavily during training, enabling the model to confidently mimic professional reporting while allegedly stealing subscription revenue and affiliate commissions from Wirecutter reviews.

Discovery evidence reportedly shows ChatGPT producing near-verbatim excerpts of copyrighted articles, with users able to bypass paywalls by requesting subsequent paragraphs. The NYT further alleges Microsoft's integration of Times-trained models across its product line generated approximately $1 trillion in market capitalization gains over the past year.

This case represents one of the most significant copyright challenges facing the AI industry, with potential implications for how tech giants develop and profit from language models trained on proprietary content.