China 🇨🇳 is considering restrictions on open access to its most powerful frontier AI models. a commentary from a Chinese state-media-linked account is proposing tiered releases for future models: • Foundational capabilities: open • Frontier capabilities: open with limits • Commercial services: closed-source • High-risk capabilities: gated and safety-tested officials have also reportedly discussed restricting foreign access to top models, including in early talks with Alibaba (no enforcement plans yet whatsoever) it's the mythos 5 incident all over again 🤧 jokes aside, everything is still in discussion, no real plans are being created yet. considering recent Chinese models releases, we can be hopeful that restrictions should not be happening anytime soon. article: