China's tightened supervision
- China appears to be easing some commercial AI rules while keeping a strong supervisory grip on the industry. - Authorities dropped a strict mandate to ensure generated content is literally 'true and accurate' but preserved heavy oversight and compliance requirements. - Analysts say this mix creates commercial space to build but keeps enforcement tight, nudging firms toward strong internal controls and fast compliance adaptation (euractiv.com) (thehansindia.com)
China has removed an explicit, literal duty for AI services to guarantee generated outputs are “true and accurate,” even as regulators tightened other controls. (mayerbrown.com) The Cyberspace Administration of China released a draft of the Interim Measures on Human‑like Interactive AI Services for consultation in April 2026, updating obligations for chatbots and virtual companions. (mayerbrown.com) Separately, China’s Measures for the Labeling of AI‑Generated Synthetic Content — issued on March 7, 2025 — began to take effect on September 1, 2025, requiring visible labels and embedded metadata for AI outputs. (gov.cn) Earlier drafts and guidance had explicitly said AI‑generated content should be “真实准确” (true and accurate), a standard critics warned was technically impossible to guarantee; the new draft drops that literal formulation. (news.qq.com) Regulators nevertheless added or preserved detailed compliance hooks: registration of human‑like AIs, ethics reviews, bans on emotional manipulation of users, and restrictions tied to national security and public order. (mayerbrown.com) Operational requirements already in force and under discussion include dual explicit/implicit labeling, mandatory metadata or watermarks, algorithm filings and sectoral security checks — all of which raise implementation costs for platforms and developers. (gov.cn) Analysts say the package opens commercial room for builders by removing an impossible literal guarantee, while keeping heavy enforcement levers that push firms to build strong internal controls and rapid compliance processes. (euractiv.com) The regulatory path evolved fast: Beijing’s Interim Measures for generative AI took effect on August 15, 2023; labelling rules followed in 2025; and targeted drafts on human‑like interactive services and ethics reviews appeared in early 2026. (whitecase.com) Companies operating in China will now navigate a regime that removes an absolute “truth” mandate but keeps routine audits, filings and potential sanctions — meaning technical teams and legal counsels must adapt within months, not years. (mayerbrown.com)
