DevelopingLegal

Unsealed Microsoft memos on AI training data surface in New York copyright suit

Decrypt, citing the New York Times, reported that unsealed filings in the Times' copyright suit against OpenAI and Microsoft show staff calling OpenAI's use of articles a giant theft of labor.

Why it matters

The filings bear on how US courts treat AI training on licensed news content, with both publishers and a model developer pressing licensing terms. This is a copyright case rather than a crypto story, so direct market impact is limited; the exposure runs to the firms' licensing costs and to how OpenAI and Microsoft source training data.

Where the sources disagree

Whether the 2023 Microsoft memos express the company's position

Internal Microsoft memos described model training as theft of unprecedented proportions and raised a doom loop

Decrypt

Microsoft says the memos were written by applied science director Brent Hecht and are not company views

Decrypt

The filings quote internal memos on AI training as theft of labor, while Microsoft says the memos were written by one employee, Brent Hecht, and do not represent company views.

Key facts

A 2023 internal Microsoft memo said millions of people would come to see large models hoovering up their work as an astonishing theft of unprecedented proportions.

Reported by Decrypt

The same 2023 Microsoft memo called large AI models a product that destroys its supply chain.

Reported by Decrypt

Microsoft employees discussed in the filings whether OpenAI's use of news articles was the largest theft of labor in human history and raised a doom loop in which models degrade on their own training material.

Reported by Decrypt

Microsoft says the 2023 memos were written by Brent Hecht, a director of applied science who also held a Northwestern University post, and do not represent company views.

Disputed figureReported by Decrypt

Microsoft CEO Satya Nadella testified that anything paywalled should be licensed by whoever wants to use it.

Reported by Decrypt

The case has forced OpenAI to preserve 20 million ChatGPT conversation logs.

20,000,000 conversation logsReported by Decrypt

Judge Sidney Stein of the Southern District of New York is weighing summary judgment motions and unsealing documents as he rules.

Reported by Decrypt

What happened

Decrypt, citing the New York Times, reported that court documents unsealed on Thursday in the New York Times' copyright suit against OpenAI and Microsoft show Microsoft employees debating whether OpenAI's use of news articles amounted to the largest theft of labor in human history. A 2023 internal Microsoft memo quoted in the filings warned that millions of people would soon read large models as 'hoovering up' their work, an 'astonishing theft of unprecedented proportions', and described large models as 'a product that destroys its supply chain'. The filings also describe a 'doom loop' concern, in which models degrade on the material they were trained on. Microsoft CEO Satya Nadella testified that anything paywalled should be licensed by whoever wants to use it, while Microsoft says the 2023 memo was written by Brent Hecht, a director of applied science who also held a Northwestern University post, and does not represent company views; OpenAI has contested the publishers' claims throughout. Judge Sidney Stein of the Southern District of New York is weighing summary judgment motions and releasing documents as he does, and the litigation has already forced OpenAI to preserve 20 million ChatGPT conversation logs. Decrypt is the only outlet the desk holds on the unsealed filings, so the account rests on a single report of documents no one else covered here.

How the story developed

  1. First report

    First report by Decrypt

    Microsoft Staff Asked If AI Scraping Was 'Largest Theft of Labor in Human History'

    Decrypt

  2. Official statement

    New York Times files copyright suit against OpenAI and Microsoft

    Eleven other publishers later joined the case.

    Decrypt

  3. Update

    Court documents unsealed as Judge Stein weighs summary judgment

    Documents were released as Judge Sidney Stein ruled on summary judgment motions in the Southern District of New York.

    Decrypt

  4. Official statement

    OpenAI ordered to preserve 20 million ChatGPT conversation logs

    The preservation duty came out of the publishers' case.

    Decrypt

  5. Update

    Decrypt reports the unsealed filings, citing the New York Times

    The report is the only version of the unsealed documents the desk has.

    Decrypt

  6. Official statement

    Microsoft says the 2023 memos do not represent company views

    Microsoft attributed the memos to applied science director Brent Hecht, who also held a Northwestern University post.

    Decrypt

Affected entities

Original sources

PublisherReportRolePublished
Decrypt
Crypto media
Microsoft Staff Asked If AI Scraping Was 'Largest Theft of Labor in Human History'Original