Microsoft knows it robbed publishers to train AI. It just doesn’t care

Ever since OpenAI’s ChatGPT chatbot and its offspring Microsoft Copilot were launched, critics have pointed out that both products — and all other generative AI tools — have been built using billions of dollars of stolen intellectual property.

Training AI models requires vast amounts of written material, especially books, newspapers, and magazine articles. OpenAI, Microsoft, and other Big AI firms have been stealing every piece of intellectual property (IP) they can find for training. With few exceptions, instead of paying the people and publishers who labored to create that intellectual property, Big AI steals it and call it “fair use.”

But just-released evidence shows that from the start, many employees and executives at Microsoft and OpenAI warned their companies that they were committing theft — and added that doing so could put newspapers and magazines out of business.

The companies ignored them and went full speed ahead.

All that and more came out last month in newly unsealed court documents filed in a lawsuit by The New York Times and others against Microsoft and OpenAI for IP theft. The filing cites memos, reports, and statements by Microsoft employees and execs that show the company was well aware of the illegality of what they were doing — and the consequences.

Here’s just one example. Back in 2023, at the dawn of genAI, an internal Microsoft memo warned: “Millions of people around the world will soon consider large models ‘hoovering up’ all their work to be an astonishing theft of unprecedented proportions.”

There’s more and worse than that in the filing. Before we get to the specifics, let’s look at the details of the suit.

A look at the Times suit

In December 2023, The New York Times sued Microsoft and OpenAI for copyright infringement and dilution of its trademark after OpenAI used millions of Times articles to train ChatGPT without the newspaper’s permission. Microsoft was included in the suit because of its ownership stake in OpenAI and because ChatGPT is the foundation for Copilot. (ChatGPT was released to the public about a year earlier, in November 2022.)

Among other complaints, the suit claims the chatbots use the paper’s reporting to allow people to read the newspaper without having to subscribe to it. It says that the paper has lost subscribers and ad revenue because of it.

 The suit adds that both ChatGPT and Copilot have created word-for-word copies of Times articles in response to people’s queries, and that the chatbots have created falsehoods (a.k.a. hallucinations) that they attribute to the Times. The suit also says its articles are one of the largest sources of training data for the chatbots, allowing the AI firms to train them more cheaply, and that Microsoft and OpenAI reap the profits while the Times loses revenue.

There’s no specific monetary value the suit asks for, although the paper has said the value of its IP is in the billions of dollars. The Times also asks in the suit that Microsoft and OpenAI destroy the datasets that include its content.

Since the suit was filed, the New York Daily News, the Center for Investigative Reporting, and seven other newspapers have joined it.

Microsoft and OpenAI execs admit theft

Employees of Microsoft and OpenAI, including high-level executives, have long agreed with the suit’s charges, even warning about the consequences of data theft before the suit was filed.

Most prominent among them is Brent Hecht, Microsoft’s director of applied science. The recent filing cites an internal memo he wrote in January 2023 calling the use of New York Times articles without the paper’s consent “an astonishing theft of unprecedented proportions” and “the largest theft of labor in human history.”

Note the date he wrote that memo: nearly a year before the Times filed its suit in December of 2023. That meant he had been warning Microsoft for at least a year before the suit that Microsoft and OpenAI were violating copyright laws and stealing IP.

Hecht had more to say. In a memo written after the suit was filed, he warned that because OpenAI and Copilot used information taken from the Times and other newspapers to answer users’ questions, people would no longer need to go to the newspapers’ web sites or subscribe to the print versions. He wrote that this “necessarily threatens the economic stability of those who create the content.”

He wasn’t alone. In June 2023, six months before the Times suite, Nick Turley, the OpenAI exec who headed the team developing ChatGPT, said that AI represented an “existential threat” to publishers.

Years before that, in 2020 before ChatGPT was even released, OpenAI’s policy director Jack Clark wrote a memo to OpenAI President Greg Brockman and OpenAI CEO Sam Altman warning 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, “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 warnings from inside Microsoft and OpenAI went on and on. They were ignored for years and continue to be ignored. And while the companies were warned that what they were doing was illegal, some execs reveled in it. When an OpenAI engineer told Brockman he was working on “a hack to get around nytimes paywall” so OpenAI could get faster access to for-pay copyrighted content, Brockman replied, “ah, nice.”

Microsoft looks the other way

Microsoft is no better. When a Times reporter asked Microsoft spokesperson Alex Haurek about Hecht’s warnings, Haurek dismissed Hecht as not being a decision maker, adding that he was hired merely “to present divergent and asymmetric perspectives.” Which were then to be ignored, of course.  

As for Microsoft CEO Satya Nadella, he backs Microsoft and OpenAI to the hilt and tries to evade responsibility for both companies’ actions. In a deposition, 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, which he says Microsoft has a right to do.

Yet for years now, he’s been aware of what OpenAI has done and how Copilot has benefited from it. Has he asked OpenAI to retrain its models on non-paywalled content?

Of course not.

Microsoft and OpenAI are flat-out wrong. The Times and the others who joined in the suit are right that their IP has been stolen. We can only hope that eventually, the courts will recognize it and finally put an end to the multi-trillion-dollar theft they’ve committed and continue to commit every day in plain sight.ComputerworldRead More