Mark Zuckerberg today published a letter, ‘Personal Superintelligence for Everyone,’ laying out Meta’s vision that artificial intelligence should help individuals pursue their own goals rather than replace human labor. The letter describes a future where AI is delivered through personal devices like AR glasses that “see what we see, hear what we hear” and interact throughout the day.

Embedded in the vision is a notable shift in tone on open source. Zuckerberg wrote that “superintelligence will raise novel safety concerns” and that Meta will be “careful about what we choose to open source.” This marks a departure from his earlier posture that Meta’s Llama models would be the industry’s most advanced open models. In a podcast last year, he had said that if there were a “qualitative change in what the thing is capable of” that made open-sourcing irresponsible, Meta would not do it.

The letter comes weeks after Meta formed Meta Superintelligence Labs and took a $14.3 billion stake in Scale AI, and amid reports that it has paused testing on its largest Llama model, Behemoth, to concentrate on a closed system. A Meta spokesperson said the company’s position on open source AI is unchanged.