Every AI Policy Ever · open in the interactive register
Published (non-binding, current)Non-binding · Technical standard
网络安全技术 生成式人工智能数据标注安全规范
Record CN-24 · GB/T 45674-2025
Recommended national standard on the security of annotating training data for generative AI. It sets requirements for annotation platforms and tools (security assessments, logging, controlled work areas), for annotation rules (separate rules for functional and security annotation, spotting risky content, incident response), for annotators (security training and exams before qualification, defined roles, with different people executing and reviewing a task) and for verification: security annotations should cover the main risks in GB/T 45654-2025 Annex A, at least 200 items per risk and 3% of the data, and a batch is discarded when over 5% fail security checks.
No English version of this instrument was found. The English title and the summary are the register's own, written from the sources below; for the exact wording use the Chinese text.
Published 25 Apr 2025 and effective 1 Nov 2025. It is a recommended 'GB/T' national standard, not a mandatory 'GB' one, and SAMR's platform lists a plan for an English-language edition. It takes its definitions of generative AI data annotation and of functional and security annotation from GB/T 45654-2025 (CN-10). It applies to organisers of annotation and can guide data requesters' acceptance checks and third-party security assessments; it also sets assessment methods for each requirement, and its annexes give annotation examples and task types.
Open the official source https://openstd.samr.gov.cn/bzgk/std/newGbInfo?hcno=407584DD0FA2BA19E62E85D3469290B0
Secondary source: https://std.samr.gov.cn/gb/search/gbDetailed?id=33D40F1160F95D92E06397BE0A0A5B93
Archived. Last attempt 2026-10-18 00:45:19 UTC
GB/T 45654-2025 China · Published (non-binding, current)
Generative AI Measures China · In force
Draft national standard on AI security capability maturity assessment China · Draft / proposed
TC260-PG-20268A (AI application security, general provisions) China · Published (non-binding, current)
TC260-PG-202610A (AI application security, health) China · Published (non-binding, current)
TC260-PG-202611A (AI application security, radio, TV and network audiovisual) China · Published (non-binding, current)
State Administration for Market Regulation / Standardization Administration of China (proposed and administered by TC260). Cybersecurity technology — Generative artificial intelligence data annotation security specification [GB/T 45674-2025]. 2025-04-25. https://openstd.samr.gov.cn/bzgk/std/newGbInfo?hcno=407584DD0FA2BA19E62E85D3469290B0. Status: Published (non-binding, current); verified 2026-10-18 (level A). Local copy archive/_runs/2026-10-14-discover-leads-2026-10-14/C008/2026-10-14/official.html retrieved 2026-10-14 13:10:40 UTC, SHA-256 e2ded420a06a7e459ba3283ee168e9810b0b1b10538a3c189e0a193904d8c483. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record CN-24.
Know of an amendment, a status change or an error? Report news or an error on GitHub; the form opens with this record filled in.