Meta CEO Mark Zuckerberg is backsliding on the company's open-source approach to AI. It's a sensible pivot.

Business Insider

31-07-2025

Of all the Big Tech CEOs, Meta's Mark Zuckerberg has been the biggest proponent of making AI more open-source. That's beginning to change now.
Ever since the generative AI boom began in 2022, Meta has made its top Llama models open-weight, which allows developers to fine-tune the model for their own use. This is pretty close to fully open-sourcing the technology and has provided developers with huge, free access to the company's best offerings in this important field.
This week, Zuckerberg began backsliding on this commitment. In a widely read essay, he suggested that Meta's powerful AI models won't always be open-sourced. On an earnings call with analysts on Wednesday, he doubled down on this idea.
Zuckerberg's change in tone
Other AI researchers have noticed this change.
"What a difference a few years makes," Nathan Lambert, senior research scientist at the Allen Institute for AI, wrote on X.
He compared a seminal Zuckerberg essay on open-source AI from 2024 with another essay the Meta CEO wrote this week on the company's approach to AI superintelligence.
"We'll need to be rigorous about mitigating these risks and careful about what we choose to open source," Zuckerberg wrote in the latest essay, published on Wednesday.
Back in 2024, he wrote, "Meta is committed to open source AI."
A 'qualitative change'
So, on Thursday morning, I asked a Meta spokesperson for a straight answer to this question: Is Meta committing to making all its most-powerful AI models open-weight in the future?
"Our position on open source AI is unchanged. We plan to continue releasing leading open source models," the spokesperson wrote in a statement. "We haven't released everything we've developed historically, and we expect to continue training a mix of open and closed models going forward."
The spokesperson also referred to a comment from Zuckerberg in 2024: "If at some point, however, there's some qualitative change in what the thing is capable of, and we feel like it's not responsible to open source it, then we won't."
It's clear to me that this "qualitative change" has happened — or at least Zuckerberg thinks it has.
This week, Zuckerberg wrote that Meta has seen early signs of AI systems improving themselves, a major requirement for this technology to reach and surpass human capabilities.
On Wednesday, the CEO told analysts that if AI models get a lot more powerful, it may not be safe to distribute them so openly, as Meta has been doing consistently with its top Llama models (so far).
"As you approach real superintelligence, I think there's a whole different set of safety concerns that I think we need to take very seriously," he said on the conference call.
Meta is investing lots of money in AI
I think there's another reason for this apparent change of heart.
Meta is spending a heck of a lot more money on AI development right now. It's paying hundreds of millions of dollars in compensation to recruit individual AI engineers and investing tens of billions of dollars more to buy and build the infrastructure needed to train and run powerful new AI models.
Turning around and giving away your best AI technology through open-weight model releases makes less sense when Meta has to make even bigger profits in the future to recoup these massive upfront investments.
"Zuckerberg's patter about openness was simply a matter of expedience. Now that he is betting the bank on his own AI project, he is singing a different tune," said Oren Etzioni, cofounder of AI startup Vercept and a veteran computer science professor at the University of Washington.
When DeepSeek's open-source models blew away Llama in late 2024 and early 2025, many AI researchers theorized that the Chinese lab used Meta's open-weight models to improve its offerings.
Zuckerberg hinted at this new competitive dynamic during a conference call with analysts late Wednesday.
"We're getting models that are so big that they're just not practical for a lot of other people to use. We kind of wrestle with whether it's productive or helpful to share that, or if that's really just primarily helping competitors," he said.
Another Zuck pivot?
There may be a simpler explanation for all this. When the world changes, you sometimes need to change with it. And Zuckerberg isn't afraid of pivoting hard. He changed the entire name of his company to pursue the metaverse, for instance.
"For as long as it makes sense and is the safe and responsible thing to do, then I think we will generally want to lean towards open source," he told The Verge last year.
"Obviously, you don't want to be locked into doing something because you said you would," he added.
Amen to that.