EU AI Act for Chinese Companies: How Article 2 Reaches You

A Shenzhen company whose AI image-generation service is used by customers in Paris is bound by the EU AI Act — even though China already regulates that same service heavily under its own rules. This is the trap for Chinese companies: having a dense domestic AI regime can create the false comfort that EU obligations are "covered." They are not. The EU AI Act is not Chinese law and never becomes Chinese law; there is no Chinese application date. But Article 2 of Regulation (EU) 2024/1689 gives the Act deliberate extraterritorial reach, and it catches Chinese providers and deployers the moment they place an AI system on the EU market or their system's output is used in the Union. China's regime and the EU's are different architectures — meeting one tells you almost nothing about the other.

---

China is a third country: the Act never becomes Chinese law

China is not an EU Member State and not part of the European Economic Area (EEA). There is no transposition route, no EEA incorporation, by which the EU AI Act enters Chinese law. So for a Chinese company the only question that matters is the extraterritorial one: does your activity bring you within Article 2?

China has a dense AI regime — but it is a different architecture

China regulates AI through a vertical, service-by-service model rather than a single horizontal risk statute. The main instruments include: the Interim Measures for the Management of Generative AI Services (in force since August 2023); the Administrative Provisions on Deep Synthesis (deepfake/synthetic-media obligations); the Administrative Provisions on Algorithm Recommendation; and the Measures for Labelling AI-Generated and Synthetic Content (issued by the Cyberspace Administration of China in March 2025), which require explicit labels and metadata-embedded markers on AI-generated content. These overlap topically with the EU AI Act — both address transparency and synthetic-content marking — but they are framed, scoped, and enforced entirely differently. There is no Chinese equivalent of Annex III "high-risk" classification or the EU conformity-assessment route.

Two regimes, no substitution

A Chinese company selling AI into the EU faces both regimes in parallel. China's measures govern services offered within China and to Chinese users; the EU AI Act governs the EU market. China's rules do not discharge the EU AI Act, and the EU AI Act does not discharge China's rules. Map your systems against both; treat the Chinese side as a matter for separate PRC counsel.

---

Article 2: the two limbs that catch Chinese companies

Article 2 sets the Regulation's scope, and it is written to reach beyond EU borders so that non-EU providers cannot undercut EU ones by staying offshore.

Placing AI on the EU market (Article 2(1)(a))

Under Article 2(1)(a), any provider that places an AI system on the EU market or puts it into service in the Union is in scope, regardless of where it is established. A Chinese company that sells, licenses, or otherwise makes its AI product available to customers in Germany, France, or Ireland is caught — incorporation in Shenzhen and servers in Hangzhou change nothing.

Output used in the Union (Article 2(1)(c))

Under Article 2(1)(c), a provider or deployer established in a third country is in scope where the output of the AI system is used in the Union. The test is where the effect lands, not where the company or server sits. A Chinese company whose model scores, ranks, or generates content used about or by people in the EU is caught on this limb even with no EU customer contract.

The authorised-representative obligation (Article 22)

A Chinese provider of a high-risk system must, before placing it on the EU market, appoint an authorised representative established in the Union under Article 22. This is a designated in-Union point of contact and accountability — a separate appointment from anything required under Chinese law.

---

Timeline: the EU dates apply to in-scope Chinese companies

There is no separate Chinese date for the EU AI Act. A Chinese company in scope through Article 2 plans against the EU statutory calendar.

The Digital Omnibus high-risk deferral

The Digital Omnibus on AI is law: Regulation (EU) 2026/1744 was approved by the European Parliament on 16 June 2026 and by the Council on 29 June 2026, published in the Official Journal on 24 July 2026, and has been in force since 27 July 2026. Stand-alone Annex III high-risk obligations apply from 2 December 2027, and product-embedded Annex I high-risk obligations from 2 August 2028.

What did not move

The Omnibus did not pause the rest of the timeline. Article 5 prohibitions and Article 4 literacy have applied since 2 February 2025; GPAI obligations under Articles 51–55 since 2 August 2025. For Chinese foundation-model and GPAI providers placing models on the EU market, the GPAI duties are live now — and models on the EU market before 2 August 2025 have until 2 August 2027 to comply under the transitional provision.

---

What in-scope Chinese companies actually have to do

The obligations are the EU obligations

Because the Act binds through Article 2 rather than any Chinese transposition, an in-scope Chinese provider or deployer faces the same role-based duties as an EU company. There is no China-lite EU file, and Chinese regulatory compliance does not stand in for it. The roles map identically: provider, deployer, importer, distributor — with importer duties at Article 23 and distributor duties at Article 24.

Article 50 transparency and synthetic-content marking

This is where Chinese providers most need to read the EU rules on their own terms. Article 50 requires, among other things, that providers mark AI-generated or manipulated synthetic content in a machine-readable way (Article 50(2)), and that deployers disclose deepfakes (Article 50(4)). All of Article 50 has applied since 2 August 2026 (generative systems already on the EU market before that date have until 2 December 2026 for the Article 50(2) marking duty; anything launched since must mark from day one). The voluntary EU Code of Practice on marking and labelling AI-generated content, finalised in June 2026, sets out how to meet it. China has its own labelling rules, but compliance with the Chinese labelling measures does not satisfy Article 50 — the technical and disclosure specifications differ, and the EU duty must be met independently for systems on the EU market.

The Article 25 role-shift

A deployer, distributor, or importer becomes a provider under Article 25 the moment it puts its own name or trademark on a high-risk system, substantially modifies it, or changes its intended purpose. A Chinese company that fine-tunes a base model into a new high-risk purpose, or rebrands a high-risk EU tool, can inherit the full provider stack without realising it.

High-risk duties and penalties

For a stand-alone Annex III high-risk system, the provider stack applies in full: risk management, data governance, technical documentation, logging, transparency, human oversight, accuracy and robustness (Articles 9–15), conformity assessment, the Article 49 registration, and provider obligations under Article 16. Penalties are the Article 99 tiers — identical figures, calculated on worldwide turnover, not EU or Chinese revenue: up to €35 million or 7% for Article 5 breaches (Article 99(3)); up to €15 million or 3% for most other obligations including high-risk provider and Article 50 transparency duties (Article 99(4)); up to €7.5 million or 1% for supplying incorrect or misleading information (Article 99(5)). For SMEs, Article 99(6) caps the fine at the lower of the fixed amount or the percentage. Separately, GPAI model providers are fined by the Commission under Article 101 (up to €15 million or 3%), not under Article 99.

---

Worked example: a Shenzhen firm offering a generative-image service in the EU

Lianyun Technology Ltd is a roughly 200-employee Shenzhen company with around ¥260 million in annual revenue. It offers an AI image- and video-generation service, and a growing share of its users are consumers and businesses in France and Germany.

Why Lianyun is in scope now

As a provider putting the service into service in the Union, Lianyun is in scope under Article 2(1)(a) today — no EU office is required, and the absence of a European entity is irrelevant. Even where users sign up without a formal EU contract, Article 2(1)(c) can catch Lianyun where the service's output (generated images and video) is used in the Union.

What bites first: Article 50, not high-risk

A general image-generation service is typically not an Annex III high-risk system by itself — but it squarely triggers Article 50 transparency. Lianyun must mark its synthetic output in a machine-readable form under Article 50(2) — a duty that has applied since 2 August 2026; because the service was already on the EU market before that date, it has until 2 December 2026 to comply — and ensure deployers can meet the Article 50(4) deepfake-disclosure duty. Its existing compliance with China's labelling measures is a starting point, not a discharge of the EU obligation.

Lianyun's checklist

• Implement Article 50(2) machine-readable marking of synthetic content to the EU specification (separate from China's labelling rules).
• Clear the Article 5 prohibited-practice gate — in particular the bans added by the Digital Omnibus, which apply from 2 December 2026: Article 5(1)(ba) (realistic intimate or sexually explicit imagery of an identifiable person without their explicit consent) and Article 5(1)(bb) (child sexual abuse material). For a provider, a reproducible outcome without adequate safeguards is caught, not only an intended purpose, so the generation filters must hold.
• If any feature crosses into an Annex III use, run the high-risk stack: appoint an Article 22 authorised representative, build the Article 11 / Annex IV documentation and Article 47 / Annex V declaration, run conformity assessment, register under Article 49.
• Note the exposure: Article 50 breaches sit in the Article 99(4) tier; as an SME, Lianyun benefits from the Article 99(6) cap at the lower of the fixed amount or the percentage.
• Keep Chinese obligations — the Interim Measures for Generative AI, the labelling and deep-synthesis rules — running in parallel; the EU file does not discharge them, and vice versa.

---

What Chinese companies should do now

Treat the two regimes as separate, not redundant

Compliance with China's generative-AI, deep-synthesis, algorithm and labelling rules does not transfer to the EU market. The EU AI Act has its own scope, its own classification (Annex III, not "service categories"), and its own transparency specification. Map each system against both regimes and meet each on its own terms.

A five-step starting checklist (EU side)

1. Map your AI systems against Article 2 to confirm where you place systems on the EU market or your output reaches the Union.
2. Classify each system: clear the Article 5 prohibited-practice gate first, then assess Article 6 / Annex III for high-risk status, and check Article 50 transparency for any generative or interactive system.
3. Assign roles — provider, deployer, importer, distributor — and run the Article 25 conversion check on anything you rebrand, fine-tune, or repurpose.
4. For high-risk systems, appoint an Article 22 authorised representative in the Union before market placement.
5. Build the compliance file: risk management, Annex IV documentation, conformity assessment, the Annex V declaration of conformity, and EU-specification synthetic-content marking — and keep your Chinese obligations running in parallel.

Two developments to monitor

Track two moving pieces for legal review: the Commission's draft high-risk classification guidelines (May 2026) and the harmonised standards still being finalised, which shape how EU high-risk duties are met, and the continuing evolution of China's vertical AI rules — which set Chinese obligations but will not displace the EU AI Act's extraterritorial reach.

---

How Confir helps Chinese companies

Confir derives your provider, deployer, importer, or distributor role and classifies your systems under Articles 5 and 6 with Annex III logic from plain-language intake, and flags Article 50 transparency duties. The engine is deterministic and rule-based — no model inference, no hallucination — so the same intake always yields the same audit-defensible finding with a human-readable explanation of which rule fired.

Confir generates the Article 11 / Annex IV technical documentation pack and the Article 47 / Annex V declaration of conformity, and supports the Article 27 Fundamental Rights Impact Assessment where it applies. The aim is to give a Chinese company an EU-market compliance file that stands on its own, separate from its obligations under Chinese law. The GPAI workflow (Articles 51–55) is partial and on the roadmap, not presented as complete. Confir is EU-hosted, at confir.eu.

Related guides

• EU AI Act extraterritorial reach for non-EU companies
• EU AI Act for US companies: Article 2 reach
• EU AI Act scope explained (Article 2)
• EU AI Act high-risk classification (Article 6)
• EU AI Act deadlines and application dates