AI and Bitstreams at the Core of China’s Patent Update

​
May 28, 2025

On April 30, 2025, the China National Intellectual Property Administration (CNIPA) released a draft amendment of the Patent Examination Guidelines (for public comment), proposing further revisions to the current Guidelines, which only came into effect in January 2024. The proposed amendments cover changes to formality examination, substantive examination, patent re-examination and invalidation and certain procedural matters related to patent examination.

In summary, with AI and bitstreams as focal points, CNIPA's latest proposal to further amend the Patent Examination Guidelines demonstrates China's ongoing efforts to refine and adapt its patent system in response to the needs arising from emerging technologies and its intention to encourage innovations and patent filings in the relevant industries.

Meanwhile, the newly drafted Guidelines are intended to clarify examination standards for AI-related and codec and streaming media patent applications. By clarifying eligibility, patentability and drafting requirements for these applications, the Guidelines may provide innovators and patent practitioners with practical guidance for patent filings and prosecution. If adopted, the draft amendments are expected to have significant impact on not only patenting activities in AI and streaming media sectors, but licensing practices in China in the future.

Highlights: AI and Bitstreams

Highlights of the amendments to substantive examination include further clarification of the criteria for eligibility and patentability for applications related to artificial intelligence (AI), and the addition of a section on the examination of applications involving bitstreams.

Notably, these amendments follow closely on the heels of the Guidance for Patent Applications for AI-Related Inventions (Trial Implementation), issued by CNIPA on December 31, 2024. This suggests that China is eager to strengthen its patent system by clarifying and adapting examination standards to address emerging technologies and some key issues of industry concerns.

AI-Related Applications: Eligibility, Ethical Rejections, and Sufficient Disclosure

In the current Guidelines, the examination of AI-related patent applications is covered under the section titled Examination of Invention Patent Applications Involving Features of Algorithms or Business Rules and Methods. Interestingly, the draft amendments propose changing the title of this section to Examination of Invention Patent Applications Relating to Artificial Intelligence, etc. This change appears to reflect an effort by CNIPA to distance AI-related inventions from business rules and methods—categories that are unpatentable—and to encourage more patent filings for AI-powered inventions in the future.

A significant proposed change is that an AI-related application may be rejected on ethical grounds. The draft amendments specify that a patent will not be granted if the invention does not comply with laws, social ethics, or public interests—especially concerning data collection, labelling management, rule setting and recommendation decision-making by AI models. Two examples are cited as unpatentable for ethical reasons: one involving an AI model that collects customers' facial information without consent and another involving an autonomous driving AI model that makes emergency decisions based on the gender and age of individuals on the road.

The draft also introduces specific requirements for sufficient disclosure of the AI algorithm or model in the application specification, addressing the "black box" nature of such systems. This may lead to more frequent objections from examiners for insufficient disclosure during examination of AI-related applications, a concern traditionally limited to chemical and pharmaceutical patent applications. Consequently, insufficient disclosure might also become a more common ground for invalidation of AI-related patents when the validity of a patent is challenged.

Bitstreams: Examination Standards and Licensing Implications in the Streaming Media Sector

The draft amendments also propose a new section outlining examination criteria for patent applications involving bitstreams, following the section on AI-related applications. This addition aims to clarify the eligibility and requirements for the specification and claims for applications for codec and streaming media technologies. It can have great impact on preparation and prosecution of patent applications for these inventions.

Regarding eligibility, the draft states that a claim involving mere bitstreams falls under rules and methods of mental activities and is thus unpatentable. To qualify for patent protection, codec and streaming media inventions must constitute a "technical solution" as defined in Article 2.2 of the Patent Law of China. This approach mirrors the examination standard for AI-related inventions.

For claims involving codec and streaming technologies, it is proposed that they may be drafted in various formats—such as a coding/decoding method for generating bitstreams, a method for storing/ transmitting the claimed bitstream, a device for implementing the claimed methods and corresponding computer-readable storage medium claims.

According to the Explanatory Notes issued with the draft amendments, these claim types are intended to align with the evolving structure of the streaming media industry and offer right holders protection for "one of the links" in the industry chain. The Explanatory Notes specifies that the aim is not to enable a right holder to assert rights across multiple links in the industry chain for licensing returns that is disproportional to its technical contribution to the industry.

This stance reveals CNIPA's view on licensing in the codec and streaming media sector. As explained in the Explanatory Notes, the industry chain of the streaming media sector is fragmented, involving multiple parties across different layers of content creation, storage, and transmission etc. Given this, CNIPA appears to oppose the licensing practice of right holders charging multiple players in different layers of the industry chain for their using the same technology. The CNIPA's position appears to be that the licensing income obtained from various players in different layers of the industry chain is disproportional to the technical contribution of the right holder to the industry.

In doing so, CNIPA appears to address the concerns of some stakeholders in the industry about the so-called "repeated charges" for licensing fees, which may be incurred to different implementors at different levels of the industry chain including hardware manufacturers and streaming platforms, as the result of the licensing practices of some right holders and patent pools in the industry. This issue has long been a topic of debate in China. CNIPA's position may have significant implications for future licensing practices in the sector.

Patenting Expertise in China

With deep knowledge of the Chinese market, we offer tailored IP advice specific to China backed by the firm's broader international expertise, aligning your patent strategies with local demands. We provide practical, market-driven insights to help you navigate China's complex IP landscape.

The Latest Thought Leadership pieces in the Content Hub

No items found.
Stay in the Know

The Patent Strategist

Get expert insights and the top patent stories delivered straight to your inbox.

Heading
This is some text inside of a div block.
This is some text inside of a div block.

Latest Insights

Showing 0 of 100
Quantum IP Strategy: Own control Points, Not Patent Counts
14 October 2026
In quantum, the strongest portfolios will not be the biggest. Stuart Malcolm sets out six moves for owning the control points the market cannot route around, and the questions investors will ask.
#
Consulting
​
Targeted, Not Transformative: The UKIPO’s Considered Course on SEPs
13 October 2026
Andrew Sharples analyses the UKIPO's response to its SEP consultation, exploring proposed changes to FRAND dispute resolution, SEP transparency and licensing guidance, and what they mean for SEP holders and implementers.
#
Consulting
​
Chinese Companies in European Patent Litigation
30 September 2026
Chinese companies are becoming increasingly active in European patent litigation, both as defendants and as patent owners enforcing their own rights. This article examines the growing role of Europe and the Unified Patent Court (UPC) in global patent disputes, highlighting key cases involving Chinese businesses and the strategic opportunities and risks that come with operating in European markets. It is the first in a series exploring the practical implications of European patent enforcement for Chinese companies.
#
Patenting
​
How to mitigate prior art traps for follow-on pharma patents
27 August 2026
Follow-on patents are increasingly important for pharmaceutical innovators. Learn how new EMA clinical trial disclosure rules create prior art risks and discover practical strategies to protect pharmaceutical IP and maximise patent value.
#
Consulting
​
Chinese companies and the UPC: from enforcement targets to strategic users
13 August 2026
Chinese companies are playing an increasingly significant role in UPC litigation, accounting for around 38% of infringement actions in 2025. While many remain defendants, a growing number are actively using the Unified Patent Court to enforce their own patent rights across Europe. The trend highlights the UPC's rising importance for Chinese businesses operating in sectors such as telecommunications, semiconductors, medical devices and renewable energy, and signals a shift in global patent enforcement strategy.
#
Patenting
​
Evaluating the environmental impact of patent AI.
31 July 2026
AI's environmental impact is rarely out of the headlines, but how does it compare to the benefits it delivers in practice? In this article, Laurence Brown examines the energy use of AI-assisted patent drafting, comparing it with traditional drafting workflows. While measuring AI's true environmental footprint is far from straightforward, the analysis suggests that AI-assisted patent drafting may already be more energy efficient, while delivering significant productivity gains for patent professionals.
#
Patenting
​
UPC appeals after three years: Early indications on appeal success rates
15 July 2026
Three years into the UPC, appeal overturn rates sit at 56% for both substantive and preliminary injunction cases, with Kodak v Fujifilm among the rulings reshaping outcomes on appeal.
#
Litigation
​
The UPC After Three Years: From Experiment to Enforcement Forum
15 July 2026
1,282 cases filed at the UPC in three years: infringement actions have surged while revocation filings stayed flat, easing fears the Court would become a European patent graveyard.
#
Litigation
​
Podcast: Building an AI-Native Law Firm
26 June 2026
On the Technically Legal podcast, EIP partner Laurence Brown discusses how Visser AI is redesigning patent workflows, letting the firm offer fixed fees without trading away speed or quality.
#
Patenting
​
Choosing your SEP battleground: UPC, Germany or UK?
07 June 2026
Choosing the right venue for SEP litigation can significantly influence the outcome of FRAND disputes. This article compares the UPC, German and UK courts, examining their approaches to injunctions, FRAND defences, rate-setting and enforcement. It highlights the strategic advantages of each jurisdiction and explains why many SEP owners are increasingly turning to the UK and UPC as part of their European enforcement strategy.
#
Litigation
​
Synthetic Lethality & AI: IP Strategy in Oncology
26 May 2026
This article explores the resurgence of synthetic lethality in precision oncology, highlighting how AI and machine learning are accelerating the discovery of new therapeutic targets. It also examines growing clinical activity, investment, and the evolving intellectual property challenges shaping this emerging field.
#
Consulting
​
Inside the Visser AI Launch Party
20 May 2026
EIP has officially launched Visser AI®, our hybrid intelligence service for patent drafting and prosecution. Watch the launch film from our event at The Boundary in Shoreditch, and discover how our team of European and US patent attorneys — working with the Patently AI platform , delivers faster, lower-cost work at the EPO, USPTO and beyond.
#
Consulting
​
Protecting Your Startup: Five Intellectual Property Mistakes to Avoid
20 May 2026
Early‑stage startups often focus on product development and fundraising, but overlooking intellectual property (IP) can seriously damage long‑term value. This article explores five common IP mistakes startups make, with a particular focus on patents. It covers the risks of filing patents without professional advice, making public disclosures too early, failing to put enforceable contracts in place with third parties, misunderstanding what patent rights actually provide, and neglecting to develop a clear IP strategy. By highlighting these pitfalls and offering practical guidance, the article helps startups protect their innovations, strengthen their commercial position, and improve their prospects with investors.
#
Consulting
​
Technology, Patents, and Sport
13 May 2026
VAR, goal-line tech, AI 3D player avatars, wearable trackers, the technology reshaping the 2026 World Cup and elite sport rests on a patent system that makes the R&D investment behind it worth taking.
#
Patenting
​
Patents or Trade Secrets? When to Patent in Deep Tech Innovation
11 May 2026
Deep tech companies face unique IP challenges due to long development cycles, high upfront investment, and complex technologies that combine products and processes. This article explores when patents, trade secrets, or a combination of both offer the strongest protection. It explains how factors such as detectability, reverse engineering risk, disclosure to third parties, and employee mobility should shape IP strategy decisions. While patents are often best suited to externally visible or licensable technologies, trade secrets can be highly effective for internal processes and know‑how that deliver performance or cost advantages.
#
Patenting
​