The lawsuit that exposed everything

Mother Jones sued OpenAI and Microsoft two years ago for training their AI on copyrighted newsroom content. Now lawsuit documents reveal how executives really think about the rules and rights of others.

The largest theft of labor ever

The lawsuit argues AI companies built their immensely valuable product via "astonishing theft of unprecedented proportions" and perhaps "the largest theft of labor in human history."

“"the largest theft of labor in human history"”

Microsoft warned of a doom loop

Microsoft's director of applied science Brent Hecht warned that AI has started a "doom loop" that will "threaten the economic foundations of its essential suppliers" and eventually make the internet infinitely worse.

“"doom loop" that will "threaten the economic foundations of its essential suppliers"”

They knew people didn't consent

An internal Microsoft document warned that millions would see AI "hoovering up" their work as theft, while acknowledging "almost no one intended for content they created to be used in this fashion, nor are they compensated."

“"almost no one intended for content they created to be used in this fashion, nor are they compensated for its use"”

AI is destroying publisher traffic

Microsoft research found that when people used AI search versus traditional search, there were 90 percent fewer clicks to news articles. Referral traffic to publishers from Google Search dropped at least 30 percent since Google introduced AI overviews.

OpenAI celebrated copying news

OpenAI president Greg Brockman excitedly told colleagues the company's products were great at spitting out publishers' content, noting ChatGPT could complete New York Times articles mid-sentence. When told about a hack to bypass the Times paywall, Brockman responded "ah nice."

“"ah nice"”

They targeted quality news content

AI companies didn't randomly grab text. They specifically zeroed in on quality content, with news articles becoming the most prevalent type in their training datasets. OpenAI even stated its goal was to "crush freshness in the domain of real-world news."

“"crush freshness…in the domain of real-world news"”

They stripped away ownership info

When creating datasets, the companies systematically removed data indicating who owned the content: copyright notices, author information, and terms of use. A Microsoft document noted most AI models are built on scraped webpages "obtained without its owners' knowledge or consent."

“"almost none of this content has been obtained with its owners' knowledge or consent"”

They filtered evidence after lawsuits

Right after lawsuits were filed, OpenAI built a special filter to stop its models from outputting content from news organizations suing it. The filter wasn't to prevent copyright infringement, but to stop plaintiffs from gathering evidence of copying.

They found 400,000 word overlaps

OpenAI analyzed 20 million ChatGPT responses searching for identical 25-word chunks in news articles. They found over 400,000 of these overlaps between ChatGPT and Mother Jones articles alone, roughly matching the New York Times count.

A prisoner's dilemma for AI

The companies are trapped in a cycle: the industry would benefit if everyone paid for content, but each individual company profits more by stealing while others pay. It's the same incentive problem driving AI companies to race ahead recklessly.

We can't trust them to self-regulate

AI executives cannot be trusted to make the right choices about stealing content or the fate of humanity. Left to their own devices, they will make wrong choices even when they know those choices are bad for them too.