Sam Altman’s OpenAI is paying publishers big money for content. That could be a lifeline — or blow up in their faces.

You’ve heard this one before: A big tech company is offering a lifeline to distressed media companies. It’s not only writing checks, but promising to help those companies adapt to seismic change.

And because this is a story we’ve heard before, we know how it will end: The media companies will once again find they got a bad deal. And they will regret their pacts with a tech giant that doesn’t care about their business — at all.

That’s the argument against a slew of deals publishers have been making with Sam Altman’s OpenAI over the past few months. Those publishers include Vox Media (my former employer), Axel Springer , Rupert Murdoch’s News Corp., Barry Diller’s Dotdash Meredith, and Laurene Powell Jobs’ magazine, The Atlantic.

The deal details haven’t been released, but some of the publishers that made them tell me the deals have the same general shape: OpenAI offers them millions of dollars in cash over the course of the deal — typically three to five years — plus some credits to help them build and operate their own products using OpenAI’s software. And the deals include promises to feature their content — along with links back to their sites — in OpenAI’s products.

In exchange, OpenAI gets the rights to use essentially everything those companies have published. That’s in part a settlement for OpenAI’s previous use of the publishers’ content, which it used to build the models that power the software company. And it’s also a license to ingest new stuff the publishers make so that OpenAI software like ChatGPT can provide answers to questions about things that are happening now.

None of this impresses Jessica Lessin, the journalist who founded and runs the tech news site The Information. She made her case forcefully in The Atlantic last month:

I also worry that this is yet another Lucy-Charlie Brown-football moment. But I figured I’d ask the people making the deals — people who are well aware of the track record for these sorts of pacts — to tell me why they think this time is different. They were happy to do so, off the record.

Here’s their argument in a nutshell:

Those other deals they did a long time ago? The ones with the likes of Google and Apple and, most particularly, Facebook? They actually have learned lessons from them.

Most crucially: Those deals required publishers to change their business — to create new formats, or make a particular kind of video or story they wouldn’t normally make, or to make more of them than they’d normally make. (The one I remember most vividly was Facebook’s live video push, which paid publishers like The New York Times to make boring videos.)

But the OpenAI deals, the publishers emphasize, are straightforward licensing deals for stuff they’re already making. Nothing bespoke. “It doesn’t change the way we operate,” one of them tells me.

Related stories

Sam Altman mulls turning OpenAI into a regular company: report

OpenAI may regret encounter with Scarlett Johansson

Using AI for work emails: which ChatGPT version is best

And that is by far the most common theme you hear when you talk to publishers about these deals. They’re something close to free money — for work that was going to get made regardless.

Which means — they say — at the end of these deals, publishers won’t have to regret investing in another defunct Big Tech project.

Another critique of these deals is that the publishers are selling their stuff too early, or for too little — and that they won’t know what the real market is for this stuff for a long time. A variation on that: Why sell this stuff now, when fellow publishers — most notably The New York Times — are suing OpenAI (and its partner Microsoft) for using their data? Why not wait and see how that pans out?

The answers to that argument aren’t quite as uniform. But if you boil them down, you can find a common theme: We don’t want to — or can’t — wait to see how this goes.

Sure, publishers tell me, these deals mean we can’t sue OpenAI for taking our stuff. But what happens if the courts — or lawmakers — end up undercutting our ability to do that anyway? It’s a crapshoot, and we don’t know when we’re going to find out the results.

And in the meantime, publishers argue, they can go ahead and strike similar deals with other big tech companies that want to use their stuff for their own AI engines. Add all of those up and this might amount to real money, right now.

But the thing I’m most worried about as someone who makes words for a living isn’t a replay of the old Facebook/Apple/Google deals publishers now regret. It would be a variation on the theme: the deals all the old media companies did with Netflix back in the early 2010s.

A reminder of how those worked: Back then, big TV networks and movie studios were delighted to sell their old titles to Netflix, which was just getting into the streaming business. The logic: We already sell our old stuff, so this is just a new business partner, and one with a big checkbook. Let’s sell them as much as we can!

The problem, as the media companies figured out too late, was that they were helping Netflix build a much better version of their own business: Why watch a show when it airs on ABC when you can wait a little while and see it on Netflix, whenever you want, without any ads?

“‘We are basically selling nuclear weapons technology to a third-world country, and now they are using it against us,” as Disney CEO Bob Iger put it.

Eventually, many of those same companies stopped selling their content to Netflix, and tried to use those shows and movies to build their own versions of Netflix — an effort that turned out to be expensive and likely too late. (Now, some of them are once again selling their stuff to Netflix because they need any money they can get.)

To me, the parallels here seem quite terrifying for publishers: What happens if consumers become used to getting answers from ChatGPT or other AI engines and stop bothering to visit the sites that generated those answers?

Some publishers I talk to believe that won’t happen — that ChatGPT users will want to read their work because their work is original and useful and because ChatGPT won’t replicate it in its entirety.

But one publisher I talked to had a more depressing, and perhaps more realistic, response to my “What if you help them become Netflix” question. “They’ve taken our movies already,” the person said.

Translation: OpenAI and other AI companies have already crawled publishers’ websites, ingested their content, and used it to train their models. That ship has sailed, the horses have left the barn, etc. All that’s left now is to accept that reality and try to make the best deal possible. And if that logic sounds less triumphant and grimly realistic — welcome to media in 2024.