Internal Microsoft memos written in January 2023 called the company's AI data practices "the largest theft of labor in human history." The company shipped Copilot anyway. That is the core of newly unsealed filings in The New York Times copyright suit against Microsoft and OpenAI, and it changes the risk calculus for every enterprise that has bet on these vendors.

The documents, unsealed last month, cite memos, reports and statements from Microsoft and OpenAI staff showing both companies knew their training-data practices were legally exposed. The suit itself was filed in December 2023. The warnings predate it by years.

What does the suit actually allege?

The New York Times sued Microsoft and OpenAI in December 2023 for copyright infringement and trademark dilution. OpenAI used millions of Times articles to train ChatGPT without permission. Microsoft is named because of its ownership stake in OpenAI and because ChatGPT underpins Copilot. ChatGPT reached the public in November 2022.

The complaint makes four specific claims:

• Chatbots let people read Times reporting without subscribing, costing the paper subscribers and ad revenue.
• ChatGPT and Copilot have reproduced Times articles word for word in response to queries.
• The chatbots have generated hallucinations falsely attributed to the Times.
• Times articles form one of the largest training-data sources, cutting the AI firms' training costs while the publisher absorbs the losses.

The suit does not state a damages figure. The paper has valued its IP in the billions of dollars. It also demands that Microsoft and OpenAI destroy the datasets containing its content. The New York Daily News, the Center for Investigative Reporting and seven other newspapers have since joined the action.

Destroying training datasets is a remedy few buyers of AI services have priced in. If a court grants it, downstream products trained on that data face an uncomfortable question about their own foundations.

What did executives know, and when?

The filing's most damaging material comes from the companies' own people. Brent Hecht, Microsoft's director of applied science, wrote an internal memo in January 2023 calling the unconsented use of Times articles "an astonishing theft of unprecedented proportions" — nearly a full year before the suit was filed. In a later memo, he warned that answering users' questions with newspaper content "necessarily threatens the economic stability of those who create the content."

Hecht was not an outlier. In June 2023, Nick Turley, the OpenAI executive heading ChatGPT development, called AI an "existential threat" to publishers. Further back, in 2020, OpenAI policy director Jack Clark wrote to President Greg Brockman and CEO Sam Altman that OpenAI's work "is going to increasingly lead to us creating systems that substitute for the labor of the people that define the 'culture' of society." He added that OpenAI "will become the symbol of how Silicon Valley is thoughtlessly stepping into other parts of life and leaving a mess on the carpet."

The paper trail also shows intent, not just awareness. When an OpenAI engineer told Brockman he was working on "a hack to get around nytimes paywall," Brockman replied: "ah, nice."

Every one of these warnings was ignored. That is the part procurement teams should sit with: the risk was documented internally, repeatedly, at executive level, and the products shipped regardless.

How has Microsoft responded?

Microsoft's public posture is deflection. Asked about Hecht's warnings, spokesperson Alex Haurek dismissed Hecht as not a decision maker, saying he was hired "to present divergent and asymmetric perspectives." Which were then ignored, apparently.

CEO Satya Nadella's deposition is more revealing. He said that if he "had been made aware that OpenAI had scraped and trained on information that was behind a paywall," he would have required OpenAI to retrain its models, and that Microsoft has the right to demand it. Yet the memos circulated for years, Copilot has benefited throughout, and Nadella has not asked OpenAI to retrain anything.

The claim of ignorance does not survive contact with the timeline. Warnings date to 2020. The suit was filed in December 2023. No retraining order has followed.

What should CIOs take from this?

The core claims remain unadjudicated. Microsoft and OpenAI maintain that training on scraped content constitutes fair use, and no court has yet ruled against them on the merits.

But the unsealed record removes the defense of inadvertence. Vendors knew. The question for enterprises is contractual: if courts eventually order dataset destruction or impose licensing costs on models built on this data, who absorbs that liability — the vendor, or the customer running Copilot across the organization?

The suit seeks no stated dollar amount yet. The Times values its IP in the billions. The products built on that data are worth trillions. That gap is where this case will be decided, and it is far from resolved.