Meta Platforms Chief Executive Officer Mark Zuckerberg signalled a policy shift on Wednesday, saying the company may no longer release the most advanced versions of its artificial-intelligence systems under open-source licences. In a public letter outlining his vision for “personal superintelligence,” Zuckerberg wrote that the technology raises “novel safety concerns” and that Meta must be “careful about what we choose to open source.” The stance marks a departure from the company’s previous pledge to make successive iterations of its Llama large-language models widely available, a strategy that had distinguished Meta from rivals OpenAI and Google DeepMind. While Zuckerberg said Meta remains committed to releasing leading open models, a spokesperson added that future work will combine both open and closed systems. The change comes as Meta accelerates spending on advanced AI. In June the company paid $14.3 billion for a 49 percent stake in training-data specialist Scale AI and consolidated its research under a new unit called Meta Superintelligence Labs. The investments, coupled with an aggressive hiring campaign, reflect Meta’s effort to keep pace in the race to develop systems capable of human-level reasoning while retaining tighter control over its most powerful tools.