Meta chief Mark Zuckerberg argued against letting powerful AI concentrate in the hands of companies and governments as the company resumed making models freely available to outside developers, positioning open-source as the answer to Chinese rivals. The social media giant released the underlying parameters of Muse Glimmer, a new "open" AI model developers can download and modify, and said it would release the weights for a version of its more powerful Muse Spark model in the coming weeks.
The move marks a return to the "open" AI strategy Meta has used to distinguish itself from rivals. Meta declined to release Muse Spark's underlying weights earlier this year, citing safety concerns, as competition increasingly focuses on how widely the most capable technology should be distributed.
Zuckerberg wrote that "widely deployed open source systems have proven more secure because more people can identify vulnerabilities," pushing back on the argument that restricting access reduces risk.
Zuckerberg positioned Meta's commitment as critical to challenging Chinese open-source models from DeepSeek and Moonshot, which he said have been getting uncomfortably close to the American frontier. His comments came as Chinese AI labs narrow the gap between their own open models and the most advanced proprietary offerings from US labs.
He also came out in favor of distillation, the technique OpenAI and Anthropic have accused Chinese rivals of using to train models through US lab outputs, saying it was important to protect the principle that you can learn from anything you observe.
Over the past year, Meta's commitment to open-weight releases had blurred as the company mixed open releases with proprietary ones. The Llama family, which expanded through several versions, helped make high-performing models available for outside developers to download and modify, giving Meta influence over the wider AI ecosystem even as rivals captured more revenue by selling access to their proprietary models.
Zuckerberg criticized rival labs for focusing on building AI for companies, governments, or other institutions, saying that if those labs lead, the balance of power will favor larger institutions over individuals. He said Meta would place the power to review the safety of its models in the hands of its independent board of directors.
