Cybersecurity Standard Practice Guide — Artificial Intelligence Application Security Guidelines: General Provisions

网络安全标准实践指南——人工智能应用安全指引总则

Record CN-20 · TC260-PG-20268A

Issued by: National Cybersecurity Standardization Technical Committee (TC260) Secretariat
Jurisdiction: China
Level · region: National · East Asia
Published / adopted: 2026-09-15
Language: Chinese
Official identifier: TC260-PG-20268A

Summary

General part of TC260's AI application security guidelines, for all sectors, based on the Cybersecurity, Data Security and Personal Information Protection Laws and the rules on generative AI, AI content labelling, algorithmic recommendation and deep synthesis. It sets principles such as ultimate human control and graded protection, and guidance for each stage from planning to decommissioning: risk identification, five-level security grading, capability maturity assessment, one-click takeover, emergency shutdown, safety tests (also for minors), guardrails, content labelling, emergency plans and annual re-testing; it encourages unconditional relief for major personal-injury accidents.

Language

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.

Safety topics

risk-based classificationhuman oversightevaluation & testingtransparency & labellingaccountability & liability

Status notes

Version v1.0-202609, released on 15 Sept 2026 with notice 网安秘字〔2026〕118号 together with sector guides for education, health, and radio, TV and network audiovisual services, to address new risks from the fast spread of AI applications. An organisation applying AI in a sector follows this guide and the sector document. The five security grades run from low to especially major security risk, and each party involved must reach a capability maturity level no lower than the application's grade; for grading and maturity assessment the guide refers to national standards on AI application security classification and grading and on AI security capability maturity assessment. Normative references include GB/T 45654-2025 (CN-10), GB 45438-2025 (CN-08), GB/T 45674-2025 (CN-24), GB/T 45652-2025 and TC260-005, AI Application Ethics Security Guidelines 1.0. Because responsibility for AI failures is hard to assign, it encourages arrangements in applications involving personal safety for immediate, unconditional humanitarian relief after major personal-injury accidents.

Sources

Open the official source https://www.tc260.org.cn/tc260/tzgg/202609/ec149159dcb246268eda751eee89967d/files/%E3%80%8A%E7%BD%91%E7%BB%9C%E5%AE%89%E5%85%A8%E6%A0%87%E5%87%86%E5%AE%9E%E8%B7%B5%E6%8C%87%E5%8D%97%E2%80%94%E2%80%94%E4%BA%BA%E5%B7%A5%E6%99%BA%E8%83%BD%E5%BA%94%E7%94%A8%E5%AE%89%E5%85%A8%E6%8C%87%E5%BC%95%20%E6%80%BB%E5%88%99%E3%80%8B.pdf.pdf

Secondary source: https://www.tc260.org.cn/tc260/tzgg/202609/ec149159dcb246268eda751eee89967d.shtml

Local copy

Archived. Last attempt 2026-10-18 00:45:04 UTC

Official source (third-party copy, not published) archive/_runs/2026-10-14-discover-leads-2026-10-14/C003/2026-10-14/official.pdf

Retrieved 2026-10-14 13:10:21 UTC

SHA-256 3c280b693b5fededb6f9a81df99c12e27fb707eaae89a60630daff7ceecf9f71

Secondary source (third-party copy, not published) archive/CN-20/2026-10-18/secondary.html

Retrieved 2026-10-18 00:13:00 UTC

SHA-256 4a4ebaf020f64af32513a71722d21882ded18e56f5fa643ecc4f82506ccf39b0

Verification

Another official source, such as an announcement by the issuer, confirms the title, date and status. checked 2026-10-18 · alternative-official-url-fetched

Recheck of 18 Oct 2026. Stored and read: the guide's PDF (tc260.org.cn, version v1.0-202609, copy of 14 Oct 2026, unchanged when the archiver checked it on 18 Oct 2026), which gives the number TC260-PG-20268A, the issuer, the month of release (2026 年 09 月) and the contents; and the release notice 网安秘字〔2026〕118号 of 15 Sept 2026 with the four guides attached (tc260.org.cn, the record's secondary source, stored on 18 Oct 2026), which gives the day of release. Stored as the record's secondary source, the notice counts only as alt-official evidence for the date, so the record is at level B; version 1 rested the date on the same notice stored as an official source in run 2026-10-14-discover-leads-2026-10-14, a file that is not among this record's stored files. The summary follows the Chinese text. No English version was found.

Second check: Independent check by Claude Code (claude-sonnet-5-5), 2026-10-18: 7 field(s) supported.

History

2026-10-18verification.level: A → B — level computed from the evidence by tools/pipeline.py2026-10-18-recheck-east-asia-1 · Claude Code (claude-opus-5-5)
2026-10-18notes: Version v1.0-202609, released on 15 Sept 2026 with notice 网安秘字〔2026〕118号 togethe → Version v1.0-202609, released on 15 Sept 2026 with notice 网安秘字〔2026〕118号 togethe — GB/T 45674-2025 linked as CN-242026-10-18-recheck-east-asia-1 · Claude Code (claude-opus-5-5)
2026-10-18related: [{"id": "CN-03", "slug": "cn-generative-ai-interim-measures"}, {"id": "CN-09", " → [{"id": "CN-03", "slug": "cn-generative-ai-interim-measures"}, {"id": "CN-09", " — added CN-24 (GB/T 45674-2025, a normative reference, clause 2) and CN-18 (the draft national standard on AI security capability maturity assessment, the kind of standard clause 7.1 d) refers to for maturity levels)2026-10-18-recheck-east-asia-1 · Claude Code (claude-opus-5-5)
2026-10-14related: [{"id": "CN-03", "slug": "cn-generative-ai-interim-measures"}, {"id": "CN-09", " → [{"id": "CN-03", "slug": "cn-generative-ai-interim-measures"}, {"id": "CN-09", " — TC260 practice guides of one series or referring to each other, as the checker of run 2026-10-14-discover-leads-2026-10-14 named them (checks/C003.json; C011's own text refers to TC260-PG-20266A)2026-10-14-discover-leads-2026-10-14 · Claude Code (claude-opus-5-5)
2026-10-14Added to the register — found in pipeline run 2026-10-14-discover-leads-2026-10-14, candidate C003 (CN/ministries)2026-10-14-discover-leads-2026-10-14 · Claude Code (claude-opus-5-5)

Related records

Generative AI Measures China · In force

AI Content Labeling Measures China · In force

Algorithmic Recommendation Provisions China · In force

Deep Synthesis Provisions China · In force

GB/T 45654-2025 China · Published (non-binding, current)

GB 45438-2025 (AI content labeling standard) China · In force

AI Safety Governance Framework 2.0 China · Superseded

TC260-PG-20269A (AI application security, education) 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)

TC260-PG-202515A (generative AI security emergency response) China · Published (non-binding, current)

GB/T 45674-2025 China · Published (non-binding, current)

Draft national standard on AI security capability maturity assessment China · Draft / proposed

Citation

National Cybersecurity Standardization Technical Committee (TC260) Secretariat. Cybersecurity Standard Practice Guide — Artificial Intelligence Application Security Guidelines: General Provisions [TC260-PG-20268A]. 2026-09-15. https://www.tc260.org.cn/tc260/tzgg/202609/ec149159dcb246268eda751eee89967d/files/%E3%80%8A%E7%BD%91%E7%BB%9C%E5%AE%89%E5%85%A8%E6%A0%87%E5%87%86%E5%AE%9E%E8%B7%B5%E6%8C%87%E5%8D%97%E2%80%94%E2%80%94%E4%BA%BA%E5%B7%A5%E6%99%BA%E8%83%BD%E5%BA%94%E7%94%A8%E5%AE%89%E5%85%A8%E6%8C%87%E5%BC%95%20%E6%80%BB%E5%88%99%E3%80%8B.pdf.pdf. Status: Published (non-binding, current); verified 2026-10-18 (level B). Local copy archive/_runs/2026-10-14-discover-leads-2026-10-14/C003/2026-10-14/official.pdf retrieved 2026-10-14 13:10:21 UTC, SHA-256 3c280b693b5fededb6f9a81df99c12e27fb707eaae89a60630daff7ceecf9f71. Every AI Policy Ever, version 1.18 (snapshot 2026-10-18), record CN-20.

Improve this record

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.