1. What's Happening
On May 19, 2026, China's national standards body (TC260) released voluntary Ethics-Safety Guidelines for AI Applications 1.0, effective July 1, including a recommended “off-switch” requirement for highly autonomous agents; two binding instruments followed on July 15 -- the CAC/NDRC/MIIT Implementation Opinions on Intelligent Agents, which define agents as a distinct regulatory category and impose a three-tier, consequence-scaled human-approval structure plus filing and compliance-testing duties in sensitive sectors, and the Interim Measures for Anthropomorphic AI Interaction Services, under which ByteDance's Doubao and Alibaba's Qwen have already shut down personalized companion-agent features for hundreds of millions of users IAPP, July 2026.
2. The Principle: Agents as Their Own Regulatory Category
China is the first jurisdiction to regulate AI agents as a category distinct from generative AI generally, rather than extending existing generative-AI rules to cover autonomous systems -- and it did so by sequencing a voluntary ethics standard two weeks ahead of binding measures, a soft-law-to-hard-law pattern regulators elsewhere may replicate as agentic deployment scales.
OUR FORECAST
Expect the EU and US state regulators to look to China's three-tier, consequence-scaled authorization structure -- and its off-switch requirement -- as a structural template once agentic AI deployment reaches comparable scale outside China. CIDAH already applies a comparable principle internally: external actions require human approval scaled to consequence, and our AI tools are calibrated to matter complexity rather than applied uniformly.
3. What We Recommend Now
- Map any China-facing or China-deployed autonomous-agent systems against the three-tier authorization and filing duties ahead of July 15 enforcement.
- For companion or anthropomorphic AI features, confirm China user data-export windows (e.g., Doubao's October 15, 2026 deadline) are honored.
This update is provided for general informational purposes and does not constitute legal advice. For guidance specific to your organization's use of AI systems, please contact Philippe Lipschutz, Adv..
This document combines the firm's advanced AI system with experienced legal expertise, under the close supervision and approval of senior attorney.