SHEIN sued Temu and lost on every count. The lesson is a checklist for anyone sending marketplace takedowns

Three years of litigation, thousands of takedown notices, a trial before the High Court — and the brand lost on every point, then was held liable to the platform it sued. Roadget Business Pte Ltd & Anor v Whaleco UK Ltd [2026] EWHC 2165 (Ch) is the most detailed court examination yet of how brand-versus-marketplace takedown campaigns actually work, and it failed for reasons that had little to do with whether copying happened. It failed on ownership paperwork and notice quality — the two things entirely within the brand’s control before anything is ever filed.

What the court actually decided

Shein’s claiming entities — Roadget Business Pte. Ltd, the group’s Singapore parent, and its UK subsidiary — sued Temu’s UK company over product photographs that appeared on Temu listings, mostly uploaded by third-party merchants clearing unsold Shein stock. Mrs Justice Bacon heard the liability trial in May 2026 and handed down judgment on 13 August 2026. The outcome, at § 361:

• Every infringement claim under the Copyright, Designs and Patents Act 1988 failed — copying (s. 17), communication to the public (s. 20), and secondary infringement (s. 23).
• Even if infringement had been shown, Temu would have kept the hosting defence under Regulation 19 of the E-Commerce Regulations (§ 361(ix)).
• Temu’s counterclaim for losses from Shein’s wrongful takedown notifications succeeded on liability, with quantum off to a further trial (§ 361(xi)).

The forensic detail is where the lessons live, because the court reconstructed the whole campaign, notice by notice.

Where the campaign came apart: ownership

Shein’s letter before action, sent 6 June 2023, asserted that Shein “owns the copyright in all photographs” on its website and demanded that Temu remove 8,036 listed URLs within three working days (§ 40). The judgment then records, from Shein’s own admissions, what was actually true that day (§ 41):

• The employee-shot photographs belonged to Guangzhou Shein, the group company that employed the photographers — not to Roadget, the entity claiming. Guangzhou Shein assigned the rights on 13 July 2023, more than a month after the letter.
• For photographs shot by suppliers and their agencies, Shein did not know whether it owned them at all.

That pattern continued under the interim injunction the court granted in September 2023. Shein notified around 770 sets of photographs for takedown; more than half turned out to be supplier or agency works outside its pleaded case, notified without any verification of the chain of title — Shein argued verification was “unworkable” and relied on warranties in its supplier contracts (§ 46). The court’s response, in February 2024, was to require Shein to demonstrate the author, the first owner and the chain of title to the claiming entity for every notified work, adding that it was “not sufficient for the claimants simply to rely on the warranties provided by their suppliers” (§ 48). When Shein’s next notices still failed to identify authors and first owners, the court confirmed that “simple assertions … as to the existence of a chain of title” were insufficient (§ 49). After that ruling, Shein never notified another supplier or agency work.

The sample work that made it to trial — photographs of a nightdress — showed why. The shots were commissioned by the supplier’s principal from a freelance photographer who did not know, and was never told, which company would use them (§ 361(i)). Until a cleanup agreement signed in June 2024 — a year into litigation — Shein held only a non-exclusive licence, no title, no right to sue (§ 361(ii)). This is not a Shein-specific pathology. It is what content ownership looks like in most multinational groups: creation happens in one entity or at suppliers, enforcement is run by another, and nobody executes assignments until a court asks for them.

What you can do yourself

Build the ownership file before the first notice, not after the first objection. For each content class — product photography, key art, promotional video, UI — record who created it (employee, agency, freelancer, supplier), which entity employed or commissioned them, what the contract says about IP, and how rights reached the entity that signs enforcement requests. Where creation sits with suppliers or freelancers, get written assignments — warranties that “the supplier has the rights” protect you in a dispute with the supplier; as SHEIN v Temu confirms, they prove nothing against a platform. Where group companies hold rights, execute inter-company assignments or an exclusive licence to the enforcing entity now; a confirmatory assignment signed mid-litigation saved Shein’s standing but not its case, and the court scrutinised every clause of it under PRC law with duelling experts.

Write notices that create knowledge. The legal trigger in every system is the platform’s knowledge, and the court set out exactly how notices fail to create it. Shein’s 8,036-URL schedule included comparison screenshots for only 287 listings; for the rest, Temu’s reviewer had to open each link himself — at which point he found one Shein image cited against 477 listings that showed completely different products, and links that led to error pages (§ 290). The court held Temu’s response — one reviewer working the list at about four minutes per listing, then a bulk removal of the remainder — entirely reasonable, and found Temu never acquired “reason to believe” the works were infringing before it removed them (§§ 293–297). A platform “is not required to accept a claimant’s assertions at face value, and may make reasonable inquiries” (§ 285). So do the reviewer’s work for them: exact source work, exact infringing URL, side-by-side evidence, the ownership chain, and the legal basis — per notice. That is also precisely the standard Article 16 of the DSA codifies for EU platforms (“sufficiently precise and adequately substantiated”), and what 17 U.S.C. § 512(c)(3) requires in the US.

Where it breaks down

The hosting defence is not a loophole; it is a description of your notice’s job. A marketplace loses Regulation 19 (and its DSA and § 512 cousins) only for specific content it knew about, or should have, and failed to remove expeditiously (§§ 299, 316). A notice that would not convince a reviewer in four minutes does not start that clock. The practical consequence cuts both ways: platforms that act promptly on notices stay protected — and brands whose notices are precise and evidenced convert the platform’s protection into an obligation to act.

Over-claiming has a price. Temu’s counterclaim succeeded because Shein notified works it had no right to notify: 15 of the 20 trial sample works went through takedown, 11 were quietly abandoned before trial, and the court dismissed the rest — so every one of those removals now sounds in damages under the cross-undertakings Shein gave to obtain its injunctions (§§ 344, 351–352). The court also rejected the argument that Temu should have somehow kept the listings alive: fashion cannot be sold without photographs, so notifying the image kills the listing, “as Shein must have known” (§ 351). The same boomerang exists outside injunctions — repeated unfounded notices ground suspension of a notifier’s access under DSA Art. 23, and knowing misrepresentation in a DMCA notice is actionable under § 512(f).

Volume is not strategy. Thousands of notices bought Shein two years of satellite litigation about the notices themselves. A hundred verified, evidenced removals beat eight thousand assertions — in speed, in credibility with the platform’s trust-and-safety team, and in what a judge later says about you.

When counsel is needed

Bring counsel in when the ownership chain crosses entities, contracts or jurisdictions — in SHEIN v Temu the dispositive documents were Chinese-law supplier and agency agreements, read by a UK court through expert evidence; when a platform stops acting on your notices and the escalation needs to put knowledge beyond argument under a named statute; and before any mass campaign or injunction, where the cost of notifying works you cannot yet prove you own is now a matter of record. The sequence that survives scrutiny is the boring one: fix title, evidence each item, notify precisely, escalate on the statute — and only then litigate, narrowly.