in

Zuckerberg Says It’s Fine to Train AI on Your Data Because It Probably Has No Value Anyway


“Individual creators or publishers tend to overestimate the value of their specific content.”

Fair Game

AI companies have been indiscriminately scraping mind-boggling amounts of content to train their AI models, a controversial practice that has led to a litany of lawsuits by copyright holders, from major record labels and newspapers to artists and authors.

But if it were up to Meta CEO Mark Zuckerberg, content creators and rights holders should just let these companies make use of their work for free.

In an interview with The Verge, Zuckerberg went as far as to argue that “individual creators or publishers tend to overestimate the value of their specific content,” a comment that will likely come as a slap in the face for those whose work is being used to create these models.

If creators were to opt out of having their content be scraped, Zuckerberg said, “then we just wouldn’t use their content.”

“It’s not like that’s going to change the outcome of this stuff that much,” he added.

His tone deaf comments are representative of an increasingly common position tech leaders in the AI space have taken. Companies like OpenAI have repeatedly argued that scraping copyrighted data falls under fair use, a doctrine that permits the limited use of copyrighted material under US law.

“When you put something out in the world, to what degree do you still get to control it and own it and license it?” Zuckerberg told The Verge. “I think that all these things are basically going to need to get relitigated and rediscussed in the AI era.”

Give It to Me

Zuckerberg is far from the first tech exec to argue that AI models should be able to rip off the work of copyright holders. OpenAI CEO Sam Altman told lawmakers during a 2023 hearing that content owners should be happy to “get significant upside benefit” from having his company’s models be trained on their work.

Microsoft AI CEO Mustafa Suleyman also argued earlier this year that any content published “on the open web” is “freeware” and should therefore be fair game — which as The Verge points out is a pretty far-out interpretation of US copyright law.

To appease content creators, AI companies have thrown money at them by signing them up for paid partnerships, something that has come to the dismay of journalists.

According to Zuckerberg, rights holders simply don’t understand that their content has little value. As such, Meta has historically shut down news outlets in countries that have proposed new rules that could have the company pay for links.

“We pay for content when it’s valuable to people,” he told The Verge. “We’re just not going to pay for content when it’s not valuable to people. I think that you’ll probably see a similar dynamic with AI.”

More on Zuckerberg: Mark Zuckerberg Goes Onstage, Gets Excited and Says Swear Word


No Priors Ep. 80 | With Andrej Karpathy from OpenAI and Tesla

Announcing fine-tuning for customization and support for new models in Azure AI 

Announcing fine-tuning for customization and support for new models in Azure AI