๐Bloomberg TechnologyโขStalecollected in 45m
China Expands Trade Secret Rules to Include AI Data
๐กNew Chinese regulations on AI data and algorithms could impact global model training and compliance strategies.
โก 30-Second TL;DR
What Changed
Trade secret rules now formally include data and algorithms
Why It Matters
This regulatory change creates significant compliance hurdles for AI companies operating in or exporting to China. It may lead to stricter data localization requirements and increased scrutiny over cross-border AI model training.
What To Do Next
Review your data governance and IP protection policies if your AI models involve cross-border data flows or operations in China.
Who should care:Enterprise & Security Teams
Key Points
- โขTrade secret rules now formally include data and algorithms
- โขRegulatory tightening aims to curb technology leaks
- โขReflects intensifying strategic competition between China and the US
๐ง Deep Insight
Web-grounded analysis with 10 cited sources.
๐ Enhanced Key Takeaways
- โขThe new regulations, effective June 1, 2026, explicitly include specific digital assets such as AI model training datasets, algorithm iteration protocols, user behavior analyses, computer programs, and source code as protectable technical information.
- โขThe scope of protection has been expanded to cover "negative knowledge," which includes intermediate results or failed experimental data, recognizing their commercial value in helping competitors avoid costly dead ends.
- โขThe updated rules introduce extraterritorial applicability, enabling Chinese authorities to take action against trade secret infringements committed abroad if they disrupt China's market order or harm domestic rights holders.
- โขThe regulations significantly strengthen administrative enforcement mechanisms, including a shift in the burden of proof for cases involving "substantially identical plus access" and increased penalties, with serious infringements potentially leading to imprisonment of up to ten years.
- โขSpecific technical confidentiality measures are now listed, such as permission grading, data desensitization, operation log retention for remote work and cross-border collaboration, and tiered access controls, to help companies protect their digital trade secrets.
๐ ๏ธ Technical Deep Dive
- Definition of Trade Secret: Under the new provisions, a trade secret is defined as "technical, operational, or other business information that is not publicly known, has commercial value, and has been subject to reasonable confidentiality measures by the right holder."
- Explicitly Protected Digital Assets: The regulations explicitly enumerate data, algorithms, computer programs, and code as technical information. This includes, but is not limited to, AI model training data, algorithm iteration protocols, user behavior analyses, and source code.
- "Negative Knowledge" Protection: Intermediate results or failed experimental data are now explicitly recognized as protectable trade secrets, provided they meet the criteria of secrecy, commercial value, and are subject to confidentiality measures.
- Reasonable Confidentiality Measures: Examples of specified measures include permission grading, data desensitization, operation log retention for remote work and cross-border collaboration, marking, classifying, isolating, encrypting, sealing for storage, restricting the scope of personnel access, and prohibiting or restricting the use, access, storage, and copying of computer, network, and storage devices.
- Improper Means of Infringement: The rules clearly specify digital circumstances that constitute improper means of infringement, such as electronic intrusion and remote data scraping.
- Burden of Proof: The new provisions ease the burden of proof for IPR holders by providing a list of preliminary evidence that can be submitted to market regulatory authorities to initiate an investigation, moving away from stricter prior requirements.
๐ฎ Future ImplicationsAI analysis grounded in cited sources
Chinese companies will likely increase investment in internal compliance systems and technical protection measures for AI-related assets.
The new regulations significantly increase the requirements for internal corporate compliance systems and introduce specific technical confidentiality measures, making robust protection crucial to avoid liability and loss of legal protection.
Foreign companies operating in China will face heightened due diligence obligations and compliance risks related to their AI data and algorithms.
The extraterritorial reach of the new rules and the explicit inclusion of digital assets mean foreign entities engaged in technical cooperation, commissioned development, or deploying localized solutions in China must redefine and document technical secret boundaries.
The new rules may lead to an increase in administrative enforcement actions and cross-border trade secret disputes involving AI technology.
The strengthened administrative enforcement powers, clearer guidance for market regulatory authorities, and extraterritorial applicability are expected to make it easier for IPR holders to pursue cases, potentially leading to more disputes.
โณ Timeline
1993
Anti-Unfair Competition Law (AUCL) adopted, establishing China's first modern trade secret statute.
1995
Provisions on Prohibiting Infringement of Trade Secrets issued, primarily addressing physical documents.
2017
Comprehensive revision of the Anti-Unfair Competition Law (AUCL).
2019
AUCL amended to expand prohibited acts and introduce burden shifting in civil litigation.
2025-08-22
Supreme People's Court ruled that core algorithm elements for AI deep learning, such as training code and databases, qualify as trade secrets.
2025-10-15
China's revised Anti-Unfair Competition Law (AUCL 2025) came into force, enhancing regulation of the digital economy.
2026-02-24
State Administration for Market Regulation (SAMR) issued the Provisions on the Protection of Trade Secrets.
2026-06-01
The new Regulation on the Protection of Trade Secrets (Provisions) came into effect, replacing the 1995 rules and explicitly covering AI data and algorithms.
๐ Sources (10)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
๐ฐ
Weekly AI Recap
Read this week's curated digest of top AI events โ
๐Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: Bloomberg Technology โ