Of all the leading tech CEOs, Meta's Mark Zuckerberg is the biggest advocate for making AI more open source. That's starting to change now.
Since the generation AI boom began in 2022, Meta has made its top llama model an open weight. This allows developers to fine-tune the model for their own use. This is pretty close to fully open-sourcing technology, providing great access to developers to the company's best services in this critical field.
This week, Zuckerberg began to turn back to this commitment. In a widely read essay, he suggested that Meta's powerful AI models are not always open source. He doubled the idea in a revenue call with analysts on Wednesday.
Zuckerberg's tone change
Other AI researchers have noticed this change.
“What a difference in a few years,” wrote Nathan Lambert, a senior research scientist at the Allen Institute of AI, wrote about X.
He compared Zuckerberg's inventive essay on open source AI in 2024 with another essay that the meta CEO wrote this week about his approach to AI Superintelligence.
“We need to be strict about mitigating these risks and be aware of those who choose to be open source,” Zuckerberg wrote in his latest essay published Wednesday.
In 2024, he wrote, “Meta is committed to open source AI.”
“Qualitative change”
So, on Thursday morning, I asked Meta's spokesman for a direct answer to this question: is Meta committed to making all the most powerful AI models openweight in the future?
“Our position on open source AI remains the same. We plan to continue releasing major open source models,” the spokesperson wrote in a statement. “We haven't released everything that we've developed historically, and we look forward to continuing to train the mix of open and closed models in the future.”
Related Stories
Business Insider tells the innovative stories you want to know
Business Insider tells the innovative stories you want to know
The spokesman also mentioned a comment from Zuckerberg in 2024.
It's clear that this “qualitative change” happened – or at least Zuckerberg thinks it has.
This week, Zuckerberg wrote that Meta saw early signs of AI systems improving itself, and that this technology is a key requirement for reaching and exceeding human capabilities.
On Wednesday, the CEO told analysts that if AI models become more powerful, it may not be safe to distribute them so openly, as Meta has been consistently doing it with the Top Llamas models (so far).
“I think there's a totally different set of safety concerns that you approach real super intelligence and we think we need to take very seriously,” he said on the conference call.
Meta invests a lot of money in AI
I think there is another reason for this obvious change in mind.
Meta is currently spending more money on AI development. They are paying hundreds of millions of dollars and investing tens of thousands more dollars to recruit individual AI engineers and buying and building the infrastructure they need to train and run powerful new AI models.
Passing AI technology through the Open-Weight Model release makes little sense if Meta has to make even greater profits in the future to recover these massive initial investments.
“Zuckerberg's pattern on openness was simply a matter of means. Now he's betting banks on his AI projects, so he's singing another song,” said Oren Etzioni, co-founder of AI Startup Vercte, a veteran computer science professor at the University of Washington.
When Deepseek's open source model blew the llamas in late 2024 and early 2025, many AI researchers theorized that Chinese labs would use open weight models of meta to improve their products.
Zuckerberg hinted at this new competitive dynamic during a conference call with analysts late Wednesday.
“We're getting a model that's so big that it's not practical for many other people to use. We're working on whether it's productive or useful to share it, or whether it's primarily supporting our competitors,” he said.
Another Zack Pivot?
All of these may have a brief explanation. When the world changes, you need to change it from time to time. And Zuckerberg is not afraid to spin violently. For example, he changed the entire name of his company to pursue Metaverse.
“As long as it makes sense, it's safe and responsible, we generally want to lean towards open source,” he told The Verge last year.
“Obviously, you said you would, so you don't want to be trapped in doing anything,” he added.
And amen.
Sign up for BI's Tech Memo Newsletter here. Please contact me by email abarr@busienssinsider.com.
