What happened

On 17 September, a batch of filings from the New York Times’ lawsuit against OpenAI and Microsoft came out from under seal. Inside were the two companies’ own internal discussions.

Nick Turley, who leads the ChatGPT team, wrote in a memo dated June 2023 that AI posed an “existential threat” to publishers. In February 2024 he added that these products “will get more and more substitutive as they get better.”

In the same batch, a Microsoft director described the scraping of content for AI as “the largest theft of labor in human history.”

The Times filed in late 2023; eleven other publishers have since joined. The judge is Sidney H. Stein, currently weighing both sides’ motions for summary judgment. What the publishers want these internal warnings to do is puncture the core of the fair-use defence: you knew all along that your product would compete with the journalism you trained it on.

Why it matters to you

You’re not a news publisher, so this looks like someone else’s fight. What makes it worth reading isn’t the verdict. It’s the date.

June 2023. ChatGPT was six months old and most brands were still asking whether the whole thing would blow over. The people building the product had already settled on the phrase “existential threat” for what it did to the people producing content — and the judgement they’d reached was substitution, not a shift in traffic mix.

That removes a familiar comfort: let’s wait until the picture is clearer. The picture was clear three years ago to the people with the best view of it. It just sat in an internal memo until a court pried it open.

There’s a second, more practical read. How this case lands will shape what the content licensing market looks like — if the fair-use defence fails, AI companies either pay for content or narrow what they take, and both routes change who decides what your content is worth. But that is years out, and nobody knows which way it goes. Anyone promising you an outcome is selling something.

What to do about it

This one doesn’t hand you a task for today. What it changes is your sense of the clock.

The work is the same as it was: how the AI engines describe you right now, whether they cite you, and whether they get you right. No verdict will answer those three for you, and whichever way it goes, the habit of asking an AI instead of clicking through isn’t going to reverse.

If you’ve spent the past three years waiting for a clearer signal before starting, this memo is probably it. It’s just that it was written three years ago.

If you want to know how the AI engines describe your brand today, we can run a round of questions first.

Further reading