AI Act vs China GenAI: what they share and how they differ

AI Act and China GenAI share 9 of 24 control themes and 5 checks. 67% of China GenAI requirements can be proven with checks AI Act already uses, 18% the other way round.

EURegulationPhasing in

AI Act

How to make an AI system compliant with the EU AI Act

33 requirementsNext Aug 2, 2027
CNImplementing rulesIn force

China GenAI

How to comply with China's generative AI rules

6 requirementsNext 2027
9/24shared control themes
5shared checks
67%of China GenAI requirements covered by AI Act evidence
18%of AI Act requirements covered by China GenAI evidence

At a glance

AI Act China GenAI
JurisdictionEuropean UnionChina
KindRegulationImplementing rules
StatusPhasing inIn force
BindingYes sameYes same
Object analysedAI system sameAI system same
ScopeProviders, deployers, importers and distributors of AI systems; providers of GPAI models.Providers of generative AI, deep synthesis or recommendation services to the public in China.
Territorial reachExtraterritorial: applies when the system is placed on the EU market or its output is used in the EU.Services available to the public in mainland China.
PenaltiesUp to €35M or 7% of worldwide turnover (prohibited practices); €15M or 3% (other obligations); €7.5M or 1% (incorrect information).Penalties under the cybersecurity, data and PIPL laws; service suspension.
Qualification axesAI Act risk level, Organisation role, General-purpose modelService type
RolesProvider, Deployer, GPAI providerService provider
Requirements336
Next milestoneAug 2, 2027, Annex I high risk (regulated products); GPAI placed on the market before Aug 20252027, Comprehensive AI Law (State Council legislative plan)

Theme by theme

requirements per theme

What they share: one piece of evidence, two frameworks

5

CodeCheckRequirements AI ActRequirements China GenAI
VER-004-01Documented data governance (collection process, bias, quality)
VER-009-02Accuracy and robustness verified and documented
VER-018-D-03AI-generated content marked as such
VER-021-D-03Operation monitoring
VER-031-G-03Documented copyright compliance policy

Differences: requirements specific to each framework

Requirements with no check serving the other framework: the extra work.

AI Act

27

EX-001
EX-003
EX-005
Technical documentation Art. 11 + Annexe IV
EX-008
EX-010
QMS Art. 17
EX-011
EX-013
EX-014
Cooperation with authorities Art. 21 + Art. 26§12
EX-016
CE marking Art. 48
EX-019
DPIA Art. 26§9
EX-020
FRIA Art. 27
EX-023
Accessibility Art. 16(l)
EX-024
EX-026
Value chain Art. 25
EX-027
Compliant use Art. 26§1
EX-028
Input data Art. 26§4

China GenAI

2

CN-01
Algorithm filing with the CAC Recommendation Art. 24

Timelines

PastSet in the textPotentialTo verify
Apr 21, 2021AI Act · Commission proposal
Mar 1, 2022China GenAI · Algorithmic recommendation provisions
Jan 10, 2023China GenAI · Deep synthesis provisions
Aug 15, 2023China GenAI · Interim measures on generative AI
Jul 12, 2024AI Act · Published in the Official Journal
Aug 1, 2024AI Act · Entry into force
Feb 2, 2025AI Act · Prohibited practices (Art. 5) and AI literacy (Art. 4)
Jul 10, 2025AI Act · GPAI Code of Practice published
Aug 2, 2025AI Act · GPAI obligations, governance, penalties, notified bodies
Sep 1, 2025China GenAI · Labelling measures and GB 45438-2025 standard
Nov 19, 2025AI Act · Digital Omnibus proposal: high-risk delay tied to standards
Jan 1, 2026China GenAI · Amended Cybersecurity Law with AI provisions
Aug 2, 2026AI Act · General application: Art. 50, sandboxes, Annex III high risk (unless Omnibus delay)
2027China GenAI · Comprehensive AI Law (State Council legislative plan)
Aug 2, 2027AI Act · Annex I high risk (regulated products); GPAI placed on the market before Aug 2025
Dec 2, 2027AI Act · Omnibus backstop for Annex III high risk
Aug 2, 2028AI Act · Omnibus backstop for Annex I high risk
Aug 2, 2030AI Act · High-risk systems of public authorities already in service (Art. 111)

Run these requirements across all your AI systems

TrustFlow inventories your systems, qualifies them under each regulation and collects evidence once for every referential.

Request a demo