Claim · 2026-09-14 · publication · Stated in source · Supported

The AI Safety Governance Framework 3.0 states that those who carry out secondary development on foundation models assess security flaws (安全缺陷) in the model and upstream components, to prevent upstream flaws spreading to downstream models and applications.

One recorded claim. The claim panel gives its qualification, the source passages with their place in each source, and how it was checked.

What it connects

AI Safety Governance Framework 3.0 (人工智能安全治理框架3.0) → states provision → Developers building on foundation models to assess security flaws of the model and upstream components

Where it appears

No collection page holds this claim. It is listed with its records in the explorer.

Claim and evidence

AI Safety Governance Framework 3.0 (人工智能安全治理框架3.0)states provision ↓Developers building on foundation models to assess security flaws of the model and upstream components

Machine-checked, not reviewed by an editorStated in sourceSupported

Supporting evidence: primary · 1 source

Provision: Required action · provision ID upstream-flaw-assessment
Subject version: 3.0 · Published 2026-09-14 · Effective date not established
Date: 2026-09-14 · publication Records dated 2026 →

Scope and limitsWhat the cited edition says, as guidance in a technical document of TC260 (a national standardisation technical committee); it is not evidence that any developer adopted, implemented or was required to follow it.

The evidence 01

primary sourceSupports the claim

人工智能安全治理框架3.0

全国网络安全标准化技术委员会 · 2026-09 · In Simplified Chinese (zh-Hans)

Archived copy, captured 2026-09-20: 人工智能安全治理框架3.0

Cite the original source

Cite the original document, not Selekoa. Check these details against it.

Source excerpt

“基于基础大模型进行二次开发的，开展模型及上游组件安全缺陷评估，防范上游缺陷向下游模型和应用扩散”

Location in source §5.1.13

How it was read Retrieved directly from the publisher

Machine-checked · 2026-10-31

Source details and check notes

Authors: 全国网络安全标准化技术委员会
Version: Dated document, retrieved on 2026-10-27 (SHA-256 cff709d8b1b2a0dfd2ff24df4d6b942b160975331e44c918e14a47d5c86da43d) directly from the publisher. Version: 《人工智能安全治理框架3.0》.pdf.pdf. Bilingual PDF attached to TC260's news item of 14 September 2026 (Chinese text followed by the publisher's English text). Quotes are from the Chinese text. English title (machine translation): “AI Safety Governance Framework 3.0”. Publisher as printed on the inspected expression: 全国网络安全标准化技术委员会 (tc260.org.cn).
Canonical address: Open the canonical source ↗
Accessed: 2026-10-27
Language: Simplified Chinese (zh-Hans)
Archive copy: Open archived copy, captured 2026-09-20: 人工智能安全治理框架3.0. An archive copy shows only what the archive retrieved on 2026-09-20. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-27.
Attribution: 全国网络安全标准化技术委员会, “人工智能安全治理框架3.0”, 2026-09.
Reuse: No open redistribution licence asserted. Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution.
Check notes: Candidate pack passage, Chinese frontier-AI institutions, third collection (research/china-labs-3). The quotation (45 words, each CJK character counted as one) matched the text retrieved for the candidate pack on 2026-10-29 and again in a fresh retrieval the same day, read directly from the publisher. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. Line-break spaces between Chinese characters and compatibility forms (⻛ for 风) in the PDF’s text layer were normalised at admission. English machine translation, not verified by a person: “Those carrying out secondary development based on foundation large models assess security flaws of the model and upstream components, to prevent upstream flaws spreading to downstream models and applications”. No human editorial review.

Status and meaning

Editorial status. Not reviewed by an editor. An exploratory record that no editor has accepted. This status does not say whether a person has read any linked passage. Each evidence record above shows how it was checked.

Relation. An exact framework edition, or another document with an exact expression, states one provision: a capability threshold or risk level and its trigger, a required safeguard or action, a decision authority (who decides, recommends, can overrule or is informed), a stated exception, or another statement of its content, such as a licence term or a prospectus statement. The passage records what the document says, not implementation, compliance or effectiveness. Provisions with the same key align across editions of one family.

Date. Release date of the cited edition, as the dataset records it; the provision states no effective date.

Record ID c3-governance-tc260-asgf-v3-upstream-flaw-assessment