Global AI governance bifurcates into US and China axes
Global AI governance splits into two axes — a China-led WAICO and a US-led voluntary pre-review framework — as each camp begins building separate norms for model weights, chip designs, data, and interoperability, initiating a bloc structure.
Weekly evidence timeline
Supporting evidence
- 2026-W30
The opening week, with signals of AI governance splitting into US and China axes clustering together. The China-led World Artificial Intelligence Cooperation Organization (WAICO) launched at the World AI Conference (WAIC) in Shanghai on 7/16, signed by 29 countries including China, Indonesia, Brazil, and Russia, with President Xi Jinping declaring its launch and setting the goal of building an international norm system separate from US-led AI governance. On the other side, the White House was reported to be poised to formally announce, before 8/1, a voluntary framework for a 30-day pre-review of frontier models with OpenAI, Anthropic, and Google (Meta excluded from the pact). China's Ministry of Commerce is reviewing rules to restrict overseas leakage of advanced AI model weights and exports of chip designs, targeting major firms such as Alibaba and ByteDance, the Financial Times reported. The EU, via the DMA on 7/16, ordered Google to grant competing AI assistants equal access to Android OS and to share search data with third-party search engines (deadline 7/27). The US, China, and EU simultaneously activated distinct AI norms, making the governance bifurcation explicitly observable.
- 2026-W31
A week in which signals of AI governance splitting along multiple axes were stamped for a second consecutive week. The European Commission began formally enforcing Article 50 transparency obligations of the AI Act on August 2 — chatbots must disclose they are AI, and deepfake video and audio must carry AI-generated markings, with fines up to 1% of global revenue for violations. On the US axis, some 1,100 employees of OpenAI, Anthropic, Google, and Meta signed an open letter on July 28 asking the US government to build an international AI pacing mechanism (signatories include OpenAI chief scientist Jakub Pachocki and Anthropic co-founders Jack Clark and Jared Kaplan). On the industry axis, NVIDIA launched an 'Open Secure AI Alliance' with 52 firms including Microsoft, IBM, SpaceX, and Hugging Face to share AI cybersecurity tools as open source, though OpenAI, Google, and Anthropic did not join. On the China axis, Moonshot AI released the full weights of the 2.8-trillion-parameter Kimi K3 on July 27 under a Modified MIT license, the largest open-weight model to date. EU enforcement, the US voluntary-regulation camp, and China's open weights simultaneously activated distinct norm paths — the governance bifurcation pattern for a second straight week, conservatively maintaining active.
- 2026-W33
A week in which AI governance advanced simultaneously along two distinct normative paths — EU enforcement and US voluntary regulation. The EU AI Act's transparency provisions (Article 50) took full effect on August 2, requiring chatbots to disclose to users that they are AI and mandating labeling of AI-generated content, with fines of up to 3% of annual worldwide revenue or 15 million euros, whichever is higher, for violations. On the US axis, the White House finalized a 30-day voluntary cybersecurity evaluation procedure for frontier models with OpenAI, Anthropic, Google, Meta, and Microsoft (run by CAISI under NIST, with open-source models excluded), but decided not to release the evaluation contents publicly. Anthropic appointed Tino Cuellar, a former California Supreme Court justice and Stanford Law School professor, as its first Chief Global Affairs Officer (CGAO), to oversee AI policy, international relations, and government cooperation. The EU's transparency-obligation enforcement and the US's non-public voluntary framework were activated at the same time along different normative paths — the governance-bifurcation pattern is reconfirmed, conservatively kept active.
Editor's note
Analysis note
A thesis that first appeared in W30. AI governance had been tracked as fragments of individual regulations (EU AI Act/DMA, US executive orders, China's generative-AI rules), but in W30 signals of these aligning into two opposing blocs — a US-led axis and a China-led axis — were stamped in a cluster within the same week. The 29-country launch of the China-led WAICO (7/16) and the White House's voluntary 30-day frontier-model pre-review framework (to be formalized before 8/1) ran in symmetric parallel, and China's review of export controls on AI model weights and chip designs plus the EU's DMA Google interoperability order overlapped, explicitly revealing each camp moving to internalize and wall off norms.
The tracking value of this thesis is whether this bifurcation is a one-off cluster of events or a structural constant in which the AI ecosystem, coupled with trade and chip export controls, splits into two norm blocs. The expansion of WAICO signatories, the actual formalization and participant scope of the White House frontier framework, the finalization of China's export-control rules, and the reactions of Google and Apple after the EU DMA deadline (7/27) are next week's tests.