"The notion that AI is so dangerous that the only safe path is an extreme concentration of power seems inherently problematic," he wrote. He also defended Meta's practice of releasing open-source AI models against criticism from some rivals over "distillation" — a technique in which one AI system is queried extensively to learn how a competing model functions — writing that "all AI models are derived from human knowledge" and that restricting such learning would put American AI development at a disadvantage. He said Meta researchers are separately using AI to help generate novel crystal structures for the company's augmented reality glasses. The manifesto arrived alongside Meta's release of an open-source AI model the company has called "Muse Glimmer," following an earlier release, "Muse Spark," weeks earlier. Zuckerberg's essay represents a continuation of a pattern of periodic public manifestos he has issued at pivotal moments for Meta's AI strategy, dating back to earlier statements on open-source AI development.