Trade Secrets 2026

CHINA – BEIJING Trends and Developments Contributed by: Ye Zhao and Liwei Jiang, Jingtian & Gongcheng

“unduly difficult” given the inherently covert nature of trade secret theft; once the claimant establishes access and substantial similarity, the burden shifts to the defendant to disprove infringement. • Rule 2 – Multiple defendants: burden further reduced: In 2023 SPC ZhiMinZhong No 642 , where evidence pointed strongly to a joint enterprise, the Court held that the claimant’s burden is “further reduced” and defendants must affirmatively prove they did not participate. • Rule 3 – Functional comparison as reverse-engineer- ing proxy: In AI fingerprint recognition case 2023 SPC ZhiMinZhong No 1503 , since the plaintiff could not access the defendant’s source code, the Court devised a point-reading test in which both products were exposed to identical test subjects and materi - als. The “high degree of consistency” in outputs – including identification, localisation, phonetic ren - dering, and even shared misidentification patterns – was held sufficient to infer that the defendant had used the plaintiff’s algorithm and training data. This methodology creates a powerful evidential route in technology cases where direct access to underlying code or technical files is practically unavailable. The combined effect of Rules 1–3 is a materially more accessible litigation path for trade secret claimants. Rights holders need no longer rely on direct evidence of misappropriation, circumstantial inference, behavioural anomalies, and product comparison can establish the prima facie case, after which the burden to rebut falls on the party with access to the relevant evidence. Trend 6: Scrutiny of the defendants’ evidence Once the claimant establishes the access-plus-sim - ilarity inference, defendants routinely respond with staged laboratory records, signed design drawings, and internal research reports. In the 2025 decisions, the SPC abandoned formalistic review and applied rigorous scrutiny across three dimensions: • Development timeline plausibility: In 2023 SPC ZhiMinZhong No 1503 , the plaintiff required 19 months and substantial human resources to develop its AI fingerprint recognition model; the defendant claimed to have completed an equiva - lent product within two months of incorporation, and without access to comparable training data

“plainly inconsistent with ordinary experience”. In No 642, the defendant alleged that a complex one- pot esterification rearrangement process was taken from first small-scale trial to commercial produc - tion in just over one month – again, held to contra - dict general scientific and industrial knowledge of development timelines. • Internal consistency of records: In 2023 SPC ZhiMinZhong No 642 , laboratory notebooks showed no corrections and recorded first-attempt success; multiple submitted versions contained discrepancies between copies and originals, and bore signs of backdating. In No 1669, work instruc - tion sheets were dated earlier than the correspond - ing design drawings they purported to implement, reversing the ordinary sequence of documentation. • Continuity of evidence chain: In 2023 SPC ZhiMin- Zhong No 2039 , the defendant admitted at trial to having destroyed all drawings, both paper and elec - tronic, of the infringing products. The Court found it “contrary to ordinary logic” to simultaneously assert independent development and acknowledge destruction of the entire technical record. Trend 7: Holistic secrecy comparison – the “salami- slicing” defence is rejected Defendants routinely attempt to disaggregate com - plex proprietary processes into individual technical features and cite separate public-domain references to show each element is already known. The SPC firmly rejected this methodology in the 2025 cases, establishing the “holistic secrecy point comparison” principle across three dimensions: • Part-public does not make the whole public: In 2022 SPC ZhiMinZhong No 445, the SPC stated: “The presence of some commonly-known tech - niques or process combinations within a body of technical information does not negate the secrecy of that information as a whole; analysis must be conducted holistically to determine whether the key elements of the overall scheme are gener - ally known to and readily obtainable by relevant persons”. In No 2880 , where the first-instance court dismissed on a point-by-point basis, the SPC reversed on the ground that the seven secrecy points together constituted a specific integrated process and must be assessed as a unit.

56 CHAMBERS.COM

Powered by