China NNN Agreements: The Complete Guide for Foreign Buyers

TL;DR

A standard Western NDA protects against disclosure of your information. In China manufacturing, the bigger risks are use (the factory makes your product for itself or for others) and circumvention (the factory sells directly to your customers or goes around you to your suppliers). An NNN agreement — non-disclosure, non-use, non-circumvention — is built for exactly those risks.
An NNN is only as good as its weakest link: the exact registered Chinese entity, execution that matches the agreement’s own effectiveness clause (signature, chop, or both), PRC governing law, a Chinese forum you can actually reach, and liquidated damages a Chinese court will enforce.
Sign before anything sensitive leaves your hands — including quotations and sampling. Files sent before signature are where every dispute starts.
An NNN binds only the companies that sign it. Register your trademark (China is strictly first-to-file), consider patents (utility models are the workhorse for OEM), and layer with a manufacturing agreement once a factory is selected.
Chinese courts do enforce these agreements, and they move far faster than Western litigation — but they are strongest where the remedy is money backed by asset freezing, and weakest at making someone do (or stop doing) something. Draft accordingly.

1. What an NNN agreement is

An NNN agreement is a contract between a foreign buyer and a Chinese supplier or manufacturer that imposes three obligations:

Non-disclosure — the Chinese party will not disclose your confidential information to others.
Non-use — the Chinese party will not use your information, designs, tooling, or business relationships for its own benefit or anyone else’s.
Non-circumvention — the Chinese party will not go around you to deal directly with your customers, distributors, suppliers, or other contacts.

It is best understood as a more comprehensive form of NDA, purpose-built for manufacturing relationships. The three risks it addresses map to the three ways foreign buyers actually get hurt in China:

IP theft — replication of designs, formulas, or product innovations.
Unauthorised production — the factory making your product, or a look-alike, for itself, for another customer, or for online platforms.
Supplier or customer circumvention — losing your customers or your supply chain to direct deals made behind your back.

A note on terminology: some documents describe the second N as “non-competition” rather than non-use, or fold a non-compete into the circumvention clause. The label matters less than the substance — what protects you is a clause that expressly prohibits the conduct you fear, not the acronym on the cover page. This guide uses “non-use” throughout, because use — not competition in the abstract — is what a supplier can actually do to you.

Two structural points explain why this instrument exists at all:

The enforcement gap is the point. A conventional NDA relies on the law of trade secrets: only information that qualifies as a trade secret now and later is protected. Much of what you share with a factory — an unpatented product concept, a customer list, pricing, sourcing relationships, a partially disclosed design — may never meet that legal standard. An NNN is instead contract-based: the contract itself defines what is protected and prohibits defined conduct, whether or not the information technically qualifies as a trade secret. Protection can also extend to information already disclosed, if identified in writing.
The forum gap decides the case. An agreement naming a foreign court and foreign law leaves you with a judgment you generally cannot convert into money in China, where the factory, its assets, and the infringing conduct all sit. Chinese courts are under no obligation to enforce a foreign judgment and, absent a bilateral judicial-assistance treaty, rarely do (see §5.5 and §9).

Why Chinese suppliers accept NNNs. The underlying concepts — don’t produce behind my back, don’t poach my customers — are familiar to Chinese manufacturers and match the norms they apply among themselves; unauthorised production and customer-poaching are precisely the disputes Chinese companies litigate against each other. Legitimate factories sign reasonable NNNs routinely. The signing process also works as a vetting tool: a factory that haggles over the damages number is bargaining; a factory that refuses to disclose its registered identity, or resists the core protections (non-use, non-circumvention, a Chinese forum, liquidated damages), is telling you how the relationship will go.

2. Why a standard NDA fails in China

The failure modes are consistent across practice:

It prohibits too little. A factory can keep your drawings perfectly confidential and still use them — to make your product for another customer, to develop its own version, or to supply a competitor. A disclosure-only clause never touches the most common harm: the factory does not leak your file, it simply builds your product itself under its own name.
It enforces in the wrong place. Foreign law plus a foreign court produces a judgment that is practically uncollectible against a mainland factory (subject to the bilateral-treaty point in §5.5). Translating the NDA into Chinese fixes neither problem.
Wrong entity. English trade names, marketplace profiles, Hong Kong affiliates, and trading companies are not the mainland manufacturer. A contract with the wrong legal entity can be impossible to enforce against the company that actually holds your files.
Trade-secrecy dependency. If protection depends on proving trade-secret status, you inherit that statute’s evidentiary burdens. Contract-based NNN obligations stand independently of trade-secret law — although the statutory trade-secret route adds remedies on top (see §10.3).
The name is not the test. An agreement called “NDA” that actually contains non-disclosure, non-use, and non-circumvention clauses under PRC law and a Chinese forum is functionally an NNN and can work fine. An agreement called “NNN” that lacks those clauses, or points disputes abroad, is worthless. What matters is the governing law, the forum, and the scope of the prohibitions — not the title on the cover.

3. The three Ns in detail

3.1 Non-use — usually the most important N

Non-use is what stops the factory from making your product, a competing product, or a look-alike. A well-drafted non-use clause expressly prohibits:

producing your products independently or for third parties;
using your designs, know-how, or tooling for other customers or for its own brand;
making look-alikes and lightly modified versions — the tweak-a-button, change-a-texture, call-it-new pattern;
reverse engineering your products or samples (a legal nuance matters here: reverse engineering of a lawfully obtained product is generally lawful under trade-secret law — which is precisely why you want a contractual prohibition; see §10.3);
registering intellectual property in its own name — trademarks, patents, and domain names (supplier trademark squatting is a recurring scenario, including squatting used offensively to file platform takedowns against you; see §10.1);
selling your products online — naming the platforms (Alibaba, Amazon, and others) so the factory cannot quietly sell where you never look;
developing a competing product from your design even if never disclosed to anyone — the core gap an ordinary NDA leaves open.

One quality check: not all documents that carry the NNN label impose the same obligations. In some, non-use exists only as an implied shadow of the confidentiality clause. Read what the clauses actually prohibit, not what the document is called.

3.2 Non-disclosure — familiar, but China has its own quirks

Standard elements: a specific definition of confidential information (product designs, CAD files, specifications, prototypes and samples, tooling and mould data, manufacturing processes, pricing and cost data, customer and supplier lists, business plans, and other non-public information disclosed during the relationship); permitted recipients; duration; and what counts as disclosure (written, verbal, electronic, incidental, and visual).

Points that deserve particular attention in China:

The group and family loophole. The signing factory may not “use” your information itself but may pass it to affiliated companies, a boss’s relative’s workshop, or family-owned entities, then claim compliance. The clause should bind the group and make the signing entity fully liable for any member’s breach. This cuts both ways: where the Chinese side also discloses materials to you, and the relationship is genuinely two-way, the same architecture protects them — which is why mutual NNNs are entirely workable when the business calls for one (see §5.7 and Q21).
Disclosure by approval process. Suppliers commonly use outside technical consultants and subcontractors in the same industry; onward disclosure should require prior written approval.
Employees. Obligations should extend to the recipient’s employees; key personnel can also sign individually, though Chinese counterparties often resist individual liability.
Marking and record-keeping. Mark correspondence and files as confidential, keep send registers and version histories — this is also the evidence that supports a statutory trade-secret claim if you ever need that route (see §10.3).
The WeChat reality. Engineers discuss design changes on WeChat, prototypes move informally, specifications evolve over months. An agreement drafted as if every disclosure arrives as a formally labelled document does not match how information actually moves. A definition that only protects marked documents leaves everything else unprotected; a bare “all information” clause may be too vague to identify what was taken. The defined information and your actual disclosure record must fit each other.

3.3 Non-circumvention — where the revenue leaks

Non-circumvention stops the factory from going around you — selling directly to your customers at a lower price, approaching your distributors, or dealing directly with your suppliers. It should expressly cover all three: your customers, your distributors, and your suppliers. Each is a real leak:

Customers are often readable off the export paperwork itself — cartons, invoices, and shipping documents tell the factory who your buyers are.
For e-commerce sellers, the factory learns your retail presence from your orders and can approach your listings with a direct-from-factory discount.
Suppliers and mills are the quiet risk: a factory that controls your supply chain can cut you off entirely — the hostage dynamic described in §11.

Circumvention is best understood as a commercial harm rather than an IP harm — the damage is lost revenue and channel control, not leakage — but in practice it is often the provision with the largest dollar impact. Its duration is usually negotiated separately and is commonly shorter than the non-use obligation (see §6).

4. When you need an NNN — and when you don’t

The test is substantive: if intellectual property or confidential information will change hands, you need NNN protection. If you are buying ordinary off-the-shelf goods and disclosing nothing of value, you do not. Everything else is staging:

Stage	Instrument	Why
Evaluating or quoting factories	NNN, signed before any sensitive disclosure	Several factories see enough to quote; only one wins. Each signs.
Joint development / idea-stage product	Product Development Agreement (PDA)	Allocates ownership in what the factory helps create — an NNN alone cannot answer “who owns the improvement?”
Factory selected, production starting	Manufacturing Agreement (MA) — ideally with the NNN clauses folded in	Quality, delivery, payment, tooling, subcontracting, defect liability.
Brands on goods, labels, packaging	Trademark registration — before the brand appears on anything sent to a factory	First-to-file. The NNN cannot bind a squatter.

Sign before disclosing. Once CAD files, specifications, prototypes, pricing, or customer information go out, disclosure cannot be undone — and waiting creates an avoidable factual dispute over what the factory received before the agreement took effect. Initial quotations should receive only non-sensitive summary information; previews, if any, should be watermarked or obscured. Leverage drops sharply once the factory has your drawings.

The manufacturing-agreement shortcut — with one trap. If the factory is already selected and a comprehensive manufacturing agreement containing non-disclosure, non-use, and non-circumvention provisions will be signed before anything sensitive is disclosed, one contract can do the job of two. Practitioners who draft these instruments for a living report that roughly half of the buyers asking them for a standalone NNN can be handled this way. The trap: if the manufacturing agreement is signed after disclosures have already happened, a boilerplate entire-agreement clause in it can be argued to supersede the NNN — including for the pre-signing evaluation period, often when the most sensitive information changes hands. Any later contract must expressly preserve claims arising from pre-signing disclosures.

When a generic NNN is not enough. The agreement must match the real transaction. Red flags that call for a tailored draft rather than a template: multiple entities involved across quoting, contracting, production, and subcontracting; disclosure includes moulds, source files, formulas, or customer and channel data; files have already been sent; enforcement needs to reach in-China evidence or assets.

5. What makes an NNN enforceable in China

5.1 The right entity — the most common fatal error

Contract only with the exact registered Chinese legal name of the mainland entity, matched to its business licence and Unified Social Credit Code (USCC). English names on invoices, websites, marketplace pages, and email signatures have no legal effect; you cannot sue a trading name — you sue the registered Chinese entity.
Verify through official sources: the National Enterprise Credit Information Publicity System (gsxt.gov.cn) for registration status, registered capital, business scope, and legal representative; China Judgements Online for litigation history; plus financial health — distressed companies are statistically more likely to misuse IP.
The three-names problem. Factories often run one name for export marketing, one Hong Kong entity for banking, and a third — the one that matters — on the mainland licence, where the machines and bank accounts are. A claim against an offshore shell leads to an empty office and a virtual address.
The Hong Kong trap. Execution by a Hong Kong trading company does not bind the mainland factory. If you buy through a Hong Kong entity, add the actual mainland manufacturer as a signatory (and vice versa).
Verify the signatory’s authority (the legal representative per the registry is safest; a sales manager can later be disowned as acting without authority) and that the person holding the chop actually sits in the building.
Map who quotes, signs, receives files, makes samples, produces, subcontracts, invoices, and holds assets — before the first disclosure.

Entity verification is the legal half of supplier due diligence. Finding and vetting the right factory in the first place is the operational half, and it starts before any contract exists. Need to find and verify the factory first? Our industrial sourcing team at Indusourcing handles machinery and auto-parts sourcing across China.

5.2 Execution: signature, chop, and the effectiveness clause

The standard best practice is the official company chop — the red, circular, star-centred seal registered with the Public Security Bureau — with the Chinese name on the chop matching the entity named in the agreement, ideally on every page. The chop is the strongest evidence that the company itself agreed and removes a litigable argument before it starts.
Which chop matters. The contract chop or the company’s official seal binds the company to an agreement of this kind; financial and departmental chops generally do not. Fake chops are a known tactic; verify the chop against the PSB registration.
As a matter of law, execution follows the agreement’s own effectiveness clause. Under Civil Code Art 490 and the SPC’s contract-book interpretation, a signature by an authorised representative can bind a company even without a chop — but where the agreement itself states that it takes effect upon signature and chop, both are required; where it states that a chop alone suffices, the chop alone controls. So “no chop means no contract” is not the law; it is one possible contractual choice. The drafting rule: make the effectiveness clause deliberate, and then execute exactly as it provides.
Signing sequence discipline: the buyer signs and dates first, sends, and waits for the fully executed agreement before disclosing anything.

5.3 Governing law: PRC law, chosen on connection and not just convenience

Choose PRC law. The choice is not merely conventional: under the closest-connection principle for contracts with foreign elements, Chinese law is the natural candidate, and it is where every practical stage of a dispute happens — evidence is in China, the defendant’s assets are in China, and any judgment or award will be executed in China. A foreign governing-law clause adds a layer where the court must be shown the content of the foreign law before applying it, which is time-consuming and expensive; if the foreign law is not proven, the court falls back to Chinese law anyway.
To be precise about the legal framework (a point of frequent confusion): PRC conflict-of-laws rules do allow parties to a contract with foreign elements to choose a foreign law, subject to mandatory provisions. Choosing it is usually a mistake for mainland manufacturing disputes — not because the choice is void, but because it buys nothing and costs proof burdens.
Note the Civil Code’s contractual freedom (Arts 464 ff.): parties may create obligations — such as standalone non-use duties — that exist independently of trade-secret law.

5.4 Language: bilingual, with the controlling version matched to the forum

Prepare the agreement in both English and Chinese so both sides genuinely understand it — and then designate one controlling version, chosen by your enforcement plan:

If you expect to enforce through Chinese courts, the Chinese version should control. Chinese courts conduct proceedings in Chinese and require Chinese translations of foreign-language evidence; the version they read is the Chinese one.
If you expect to enforce through arbitration with English as the language of the arbitration, the English version can control. CIETAC and other Chinese arbitration institutions can conduct proceedings in English, and an English-controlling text matches that procedure.
Making both versions equally controlling invites a text-discrepancy fight; pick one. And the Chinese text should be drafted in Chinese from the start by a Chinese-speaking lawyer, not translated after the fact from an English original — a weak Chinese text is not rescued by a controlling-language clause.

One trade-off worth naming: arbitration in English buys procedural comfort, but it is slower to first coercive action than Chinese court litigation — a point developed in §5.5.

5.5 Dispute resolution: court or arbitration — decided by speed and enforcement

This is the most consequential drafting choice, and the two routes serve different needs:

Chinese courts — the default for speed. For a dispute with a mainland manufacturer, litigation in the Chinese courts offers:

Fast coercive response. Pre-suit and during proceedings, the court can freeze the factory’s bank accounts and seize assets (asset preservation) — often before the factory even answers the claim. For urgent situations, pre-suit evidence preservation is also available. This speed to real coercive leverage is the single strongest argument for the courts.
Conduct-type relief. Orders to stop infringing activity (injunction-type relief) are available from courts; arbitral tribunals have historically been weak on relief that requires someone to do or stop doing something.
Speed and cost. Judgment typically arrives far faster than comparable Western litigation, and costs are modest by Western standards.
Enforcement at home. A Chinese judgment executes against Chinese assets through the court’s own enforcement machinery.

Arbitration — when the case for it is structural. Arbitration (typically CIETAC in the mainland; HKIAC or SIAC in the region) has real advantages in the right configuration:

English proceedings and a forum that international counterparties perceive as neutral.
A New York Convention award — enforceable in the Convention’s 170-plus contracting states (China acceded in 1987, with a reciprocity reservation) — which matters if the counterparty has assets outside China or you may need recognition abroad.
Finality (no appeal) and confidentiality.
The emergency-arbitrator mechanism now available in major rules allows interim measures — including evidence preservation — to be sought before the tribunal is constituted, so arbitration is no longer helpless in urgent scenarios. In practice, though, this route still responds more slowly than a Chinese court’s pre-suit preservation.

Two configurations steer the choice toward arbitration: counterparties that refuse local-court jurisdiction (state-owned enterprises often accept arbitration but not a local court), and multi-party or multi-jurisdiction disputes that need a single consolidated forum.

The enforcement asymmetry that should shape your drafting. Chinese judges have near-unlimited coercive power over money — account freezes, asset seizure, judgment enforcement — and far weaker tools for conduct: a judgment ordering someone to do or stop doing something is hard to police, and judges rarely issue orders they know will be ignored. The drafting consequence inverts common-law instinct: rather than treating an injunction as the ultimate remedy for unquantifiable harm, convert every feared breach into a specific, pre-agreed sum of money, so that you never need to rely on an injunction at all.

Forum-clause hygiene. Clauses copied from US templates, references to courts with no connection to the dispute, wrongly named arbitration institutions, and competing court-plus-arbitration clauses are all defective. Under the revised Arbitration Law (effective 1 March 2026), an arbitration clause missing essential terms may be void if the parties cannot cure it. Venue in litigation generally follows the defendant’s registration or place of business, or the place of contract performance — for manufacturing disputes, typically the factory’s city. Specialised IP courts operate in Beijing, Shanghai, Guangzhou and Hainan, with IP tribunals in many more cities.

Foreign judgments — the bilateral-treaty test. Whether a foreign court judgment can be recognised and enforced in China depends, as a practical matter, on whether China and the judgment country are party to a bilateral agreement on judicial assistance that covers it. Where such a treaty exists, enforcement is possible; where none exists, recognition depends on case-by-case reciprocity and the realistic prospects are poor. The drafting conclusion is the same either way: design the forum for enforcement in China from the start, and where no treaty covers your counterparty, arbitration under the New York Convention is the safer cross-border instrument.

5.6 Liquidated damages — the clause that does the real work

Why they matter. Proving actual damages from IP misuse is the hardest part of any IP litigation — how do you prove sales lost to a clone sold through a shell company? The Civil Code (Art 585, within the breach-remedies framework of Arts 577–588) lets parties pre-agree damages, and Chinese courts enforce and even encourage such clauses. They do three jobs: deterrence (the factory knows the exact price of breach in advance), simplification (no need to prove actual loss), and leverage (a demand letter citing a specific number carries weight).

The adjustment rule — and the 30% line. A court may, on request, increase an amount below actual loss or reduce one that is excessively high. The SPC’s 2023 contract-book interpretation (Art 65) supplies the working benchmark: agreed damages exceeding actual loss by more than roughly 30% will generally be treated as excessive. Practical consequence: calibrate liquidated damages to within about 30% of your realistic loss — a number with a defensible relationship to the harm (lost profits, market damage, development cost), assessed against the transaction, the parties’ performance, and fault. Fantasy numbers are not deterrents; they are reductions waiting to happen.

Where the number earns its keep: asset preservation. The agreed amount is the anchor for a pre-suit or mid-case asset-preservation application — one of the clause’s core functions and the source of most of its real-world deterrence. Security rules: security for preservation during litigation is generally capped at 30% of the preserved amount; pre-suit preservation normally requires full security (with judicial discretion in special cases); insurance and guarantee-company bonds are available.

Additional remedy-drafting points:

State whether liquidated damages are in addition to actual damages exceeding them — and decide deliberately.
Prevailing-party attorney’s fees and investigation-cost recovery discourage frivolous defence.
Connect each remedy to a defined breach and an evidence source — the file, the unauthorised act, the recipient, the protected customer, or the filing that triggers it.
IP clawback clause: any trademark, patent, copyright, or domain the counterparty registers without written approval is transferred back to you. Cheap to draft; directly answers the supplier-squatting scenario.
Non-monetary remedies exist but should not carry the agreement: urgent conduct preservation is available, and courts generally support return, deletion, and takedown — but none is automatic, and the money-versus-conduct enforcement asymmetry means money remains the reliable lever.

5.7 Scope that matches reality

Define the permitted purpose — always. The agreement must state what the information may be used for (quotation, feasibility review, sampling, or the identified order); anything beyond that purpose is a breach. An agreement that prohibits “use” without defining the permitted use has no baseline: there is nothing for “outside the purpose” to mean. This is not merely preferable but necessary — the permitted-purpose definition is what makes the non-use clause enforceable.
Cover pre-signing disclosures. Files already sent during last month’s quoting must be expressly covered. Civil Code Article 501 imposes confidentiality duties during negotiation even where no contract results — but you still must prove what was shared, with whom, under what expectation, and how it was misused. A later NNN can acknowledge prior delivery and regulate continued holding, future use, return, and deletion; it does not automatically convert past conduct into contractual breach.
Unpack duration into four distinct clocks: (1) the agreement term; (2) the disclosure window; (3) the contractual duty; (4) how long the information remains valuable. A copied-in “two-year term” collapses all four into one clock and can gut the protection. See §6.
Improvements and new IP — follow the statutory default, then contract around it. Under PRC law, technical improvements made by the Chinese party belong to the Chinese party as the legal default. The contract should then do two things with that default: (a) grant the foreign buyer a free, continuing licence to use the improvement, and (b) prohibit the Chinese party from using the improvement to produce products that compete with yours. Deeper allocation of jointly developed IP belongs in the product development or manufacturing agreement; but the licence-plus-no-competition pair belongs in the NNN itself, because it is what closes the improve-and-compete loophole from day one.

6. Duration

Term design should distinguish the layers rather than impose one clock:

The core obligations are permanent. The non-disclosure, non-use, and non-circumvention obligations survive termination of the agreement indefinitely. This matches their contractual nature: they do not depend on the information remaining a trade secret, so there is no doctrinal reason for them to expire with the agreement.
Document-delivery and supply provisions can carry a fixed term — three to five years is a workable range for the operational parts of the relationship.
Anchor the surviving obligations to the last disclosure where a fixed period is negotiated. If the Chinese side pushes back on an indefinite tail (they often do, and one-year counters are common in the market), a compromise measured from the last protected disclosure rather than from signing is more sensible than everlasting protection for information whose commercial value may fade sooner.
Non-circumvention is commercially negotiable and is commonly set at two to three years post-termination.

In short: permanent N-D-U-N-C obligations; a 3–5 year term for the document and supply mechanics; and if the counterparty insists on a tail, measure it from the last disclosure.

7. Subcontractors, affiliates, and the whole supply chain

Subcontracting is pervasive in Chinese manufacturing — the factory that signs is often not the factory that cuts, dyes, moulds, or sews. Your contract binds the company that signed it; the company that dyes the fabric or cuts the tool signed nothing.

Subcontracting is permitted where the business needs it — an outright ban is neither realistic nor necessary — but it must be structured:

The subcontractor signs its own NNN with the manufacturer. Flow-down language requires the manufacturer to impose equivalent written NNN obligations on any subcontractor before onward disclosure, with prior written approval for material subcontracting.
Joint and several liability. If a subcontractor breaches, the Chinese manufacturer and the subcontractor answer jointly and severally to you. Without this, the factory escapes by saying “our subcontractor did it” — the most common deflection in practice.
Direct binding for critical players. Where another company will receive particularly sensitive information, hold critical tooling, or perform an important part of the work, consider binding that company directly with its own NNN rather than relying on the chain.
Beyond the factory walls. Where a relationship deteriorates, the extreme version of supply-chain risk is real: factories have turned mills and upstream suppliers against buyers (§11). NNN coverage of the wider chain — mills, key upstream suppliers — is the structural answer.
Information splitting. Share only what production strictly requires, or split parts across independent factories so no single supplier holds the whole; buy the critical material or process yourself if feasible.

8. Tooling and moulds

Moulds are where ownership, money, and leverage collide, and they generate a recurring failure pattern. Five distinct legal questions must not be blurred:

Ownership — who owns the tooling, decided by who paid and what the agreement says, not by whose product it makes;
Possession and custody — who holds it, and where (a factory holding your paid moulds has enormous leverage; consider neutral warehousing on termination);
Permitted use — what it may be used for: nothing beyond your orders;
Return — what happens on termination: timing, costs, export logistics — and what stops production that should have ended;
Design files — CAD, patterns, specifications, and test data are separate from the physical tool; ownership of the tool and of the file that generated it are separate questions.

Four things to settle in writing before a tool is cut: who owns it; who holds it and where; what happens on termination; who owns the design files. Paying for the mould settles almost nothing without the paper.

The stakes are not hypothetical. Documented scenarios include factories demanding double-digit price increases while holding the buyer’s moulds, designs, and customer information — with the replacement-mould shutdown threatening the buyer’s survival — and factories asserting, in writing, that the buyer’s fabrics and suppliers had become the factory’s own property.

9. If a dispute has already started

Evidence first, accusation second. Before sending anything, freeze the record: original emails and chat threads, file versions, access logs, recipients, marketplace listings, samples. Isolated screenshots are weak. Preserve the counterparty’s own admissions — a factory’s written claim that your fabrics and suppliers “are our company property” is the kind of statement that decides cases.
The demand letter should be surgical, not generic. Specify which materials are protected, when they were disclosed, exactly what the supplier did, which clause is breached, what must stop, and what must be delivered: compensation, return, deletion, withdrawal of an IP filing, takedown — with a deadline. Generic “you violate trade secrets” letters underperform; specific ones settle, because many Chinese companies will settle rather than face litigation.
Emergency NNNs for the rest of the chain. When a relationship blows up, extend NNN protection to the rest of your supply chain within days — before those relationships can be turned against you. Most importers think an NNN is something you sign at the start of a relationship; it is also the emergency brake.
Don’t pay the hostage. Where a factory withholds goods, moulds, or supply-chain cooperation to extract a disputed balance, release-milestone payment terms in an enforceable contract are the structural fix; paying in full under threat funds the next hostage-taking.
Move for preservation early. Freezing the factory’s accounts often brings it to the table. Litigation costs are lower than folklore suggests — full suit-plus-appeal engagements in the low tens of thousands of RMB are reported.
The statutory routes alongside contract: trade-secret civil claims (AUCL 2025 revision, Art 10); administrative complaints to SAMR; criminal referral for serious misappropriation; platform takedowns under the E-Commerce Law for listings; customs recordal and enforcement for exports.

10. The NNN is one layer: the rest of the stack

An NNN binds only its signatories. It creates no rights against anyone else — the trading company you never contracted with, the competitor two booths down at the trade show, the squatter. Practice is unanimous on the point: register your IP in China, early.

10.1 Trademarks — before the brand appears on anything sent to a factory

China is first-to-file (Trademark Law Art 31): filing date beats use, anywhere in the world. Registration typically issues in about a year (10-year term, renewable; three years of non-use exposes the mark to cancellation — keep packaging and sales records).
File in both Latin letters and Chinese characters — transliteration squatting is a standard attack — and in all relevant classes (each class is a separate filing; market rates roughly USD 300–800 per class including fees).
Never engage agents promising unusually fast approval; the timeline is what it is.
The classic horror story is the subcontractor that registered the buyer’s trademark: SMEs have paid substantial sums to buy back marks they could have registered for under EUR 300 before disclosure. Buyers have also discovered their own factory registering their trademarks while producing look-alikes for their competitors.
Beyond the NNN: Trademark Law Art 15 bars an agent or representative from registering the principal’s mark without authorisation — a direct statutory hook against a signing supplier’s squatting; Art 44 provides invalidation of bad-faith registrations; and a registered mark unlocks customs recordal and platform takedowns under the E-Commerce Law.

For the full filing strategy — classes, the Chinese-character mark, bad-faith oppositions, and customs recordal — see the companion guide, Trademark Protection in China: First-to-File, Squatting, and What to Do Before You Launch.

10.2 Patents — utility models are the OEM workhorse

Type	Term	Market cost	Grant time	Best for
Invention	20 yrs	~USD 2,000–5,000	18–24+ months	Novel mechanisms/formulations
Utility model	10 yrs	~USD 500–1,500	6–12 months	Most OEM products — best cost/protection ratio
Design	15 yrs (since 2021)	~USD 400–1,000	3–6 months	Shape/aesthetics

Overlapping strategy: file a utility model on the core functional feature plus a design patent on the distinctive shape before production. Note the timing trap: once a patent application publishes, the published information has left the secrecy zone. Registered IP does not depend on secrecy — which is exactly why the registration layer complements the contract layer.

10.3 China’s trade-secret law — the statutory layer beside your contract

AUCL (2025 revision, effective 15 October 2025) — trade secrets sit in Art 10 (renumbered from Art 9 by the 2025 revision). Art 10 expressly covers obtaining by improper means (theft, bribery, fraud, coercion, electronic intrusion) and disclosure and use in breach of confidentiality duties; circumvention-based acquisition is also caught. New Art 39 shifts the evidentiary burden in trade-secret civil cases: once the right holder produces preliminary evidence, the burden moves to the defendant to show the information is not a trade secret.
SAMR Trade Secret Protection Provisions (effective 1 June 2026): Art 9 confirms that a signed confidentiality agreement is an appropriate protective measure; Art 15 confirms that reverse engineering of a lawfully obtained product is generally lawful. Trade-secret protection therefore thins once your product is on the market — which is precisely when the contractual non-use clause keeps working. The regulation is the safety net; the NNN is the harness you put on before you climb.
The tripartite trade-secret test: not publicly known + commercial value + reasonable confidentiality measures — the last element is in your control, and your signed NNN plus marked documents, restricted circulation, and access logs are the evidence.
SPC case (2022), docket no. (2022) Zui Gao Fa Zhi Min Zhong No. 1981: agreed confidentiality terms can affect proof of trade-secret status and shift evidentiary burdens — the defined information and the actual disclosure record must fit each other.
Reported win rates for foreign plaintiffs in Chinese IP litigation are encouraging — figures circulating in practice include roughly 68% overall and higher in some specialised courts — but treat any win-rate statistic with caution unless you can check its source and year before relying on it in a decision.

11. Enforcement reality

Chinese courts do enforce NNN obligations, faster and cheaper than Western litigation folklore suggests. Structural advantages: early asset preservation; specialised IP courts and tribunals; high-volume, professionally run dockets. Foreign arbitral awards are enforceable under the New York Convention, and large foreign awards have been recognised and enforced by mainland courts — a nine-figure USD SCC award enforced by the Liaocheng Intermediate Court in 2020 is the frequently cited example. The recurring war stories share a shape:

The long relationship that wasn’t protection. A Western fashion brand used the same garment factory for seven years and millions of dollars of orders. The factory contacted the brand’s own customer directly, and when confronted replied in writing that the fabrics and the suppliers were the factory’s property — only the finished products were. With goods in transit as leverage, the response combined a supply-chain-wide NNN extension and a lawsuit; total litigation cost including appeal was reported at around RMB 30,000 (roughly USD 4,200). Loyalty is not protection — leverage comes from paper and structure.
The mould hostage. A factory holding the buyer’s moulds, designs, and customer information demanded a double-digit price increase; the alternative — replacing the moulds — meant a shutdown that nearly bankrupted the buyer.
The wrong signature. An agreement signed by a Hong Kong sourcing company while the mainland factory holding the CAD files signed nothing — and later opened a second production line for a competitor.
The trade-show founder. Two years building a product, an NDA drafted by a lawyer and signed by both sides — and six months later, the product was on another company’s table under a different name. The paper had been written for the wrong country.

These circulate as practitioner war stories rather than court records; treat them as illustrations of failure modes, not citations.

12. Costs, timelines, and market rates

Item	Range	Note
Lawyer-drafted bilingual NNN	USD 800–2,500 typical; USD 1,500–5,000 complex/multi-party	market ranges, not quotes
Drafting time	2–3 days (established template + verification) to 2–3 weeks (complex, negotiated)	—
Adding a factory to an existing NNN	a fraction of the first engagement; days not weeks	—
Chinese lawsuit including appeal	reported engagements in the low tens of thousands of RMB	anecdotal
Trademark	USD 300–800 per class; 6–12 months	—
Patents	USD 400–5,000 by type (see §10.2)	—
Free or cheap templates	USD 0–500	workable only where sensitivity is low, the entity is single and verified, and the scope is tightly limited

A useful cost anchor: a single injection mould routinely costs more than the manufacturing agreement; a first production deposit costs several times more; a season of defective or late inventory costs more than all of them together.

13. The mistakes checklist

Using a foreign-law agreement with a foreign forum. 2. Wrong entity (English trade name / Hong Kong shell / trading company). 3. Execution that doesn’t match the effectiveness clause — or the wrong chop (financial/departmental/fake). 4. Disclosing before execution, including “harmless” previews. 5. No liquidated damages — or numbers beyond the ~30% benchmark that invite reduction. 6. Vague confidential-information definition, or marking-only protection. 7. Missing non-use or non-circumvention entirely. 8. No subcontractor flow-down, no joint-and-several liability. 9. No affiliate/group liability. 10. No tooling and mould terms. 11. No improvement terms (free licence to buyer + no-compete on the improvement). 12. No IP clawback for unauthorised filings. 13. An entire-agreement clause in a later manufacturing agreement wiping out pre-signing claims. 14. No trademark registration before disclosure. 15. No supplier due diligence (registry, litigation history, financial health). 16. Signing without verifying signatory authority. 17. A mutual structure where the business is actually one-way (or vice versa). 18. Relying on machine translation. 19. No permitted-purpose definition. 20. Treating the NNN as a substitute for a manufacturing agreement or product development agreement.

About the author

Brad Shu is a partner at DHH Law Firm’s Shanghai office and has practised PRC law for over twenty years, including nearly a decade between the Beijing offices of US firms Squire Sanders and Morrison Foerster and leading local firm Jingtian & Gongcheng. He drafts and advises on NNN, OEM, and manufacturing agreements for foreign buyers sourcing from China.

Email: [email protected]
Firm: DHH Law Firm, Shanghai Office
Profile: About Brad Shu

The rest of this guide: the FAQ series

This guide is the core text of the NNN agreements topic. The sixty-two questions buyers actually ask are answered across six companion articles in this topic:

Part	Article	Questions
FAQ 1	China NNN Agreements FAQ — Basics, Definitions, and What You Are Signing	Q1–Q10
FAQ 2	China NNN Agreements FAQ — Whether and When You Need One	Q11–Q20
FAQ 3	China NNN Agreements FAQ — Drafting the Agreement	Q21–Q30
FAQ 4	China NNN Agreements FAQ — Language, Governing Law, Execution, and Verification	Q31–Q40
FAQ 5	China NNN Agreements FAQ — Money, Damages, Disputes, and Enforcement	Q41–Q50
FAQ 6	China NNN Agreements FAQ — Scenarios, Coverage Gaps, and Common Situations	Q51–Q62

Each FAQ article also carries a short introduction on its theme; the six are best read as one continuous question-by-question companion to this guide.