In the era of AI, your intellectual property strategy should be replaced
Source: IPRdaily Chinese website (iprdaily. cn)
GIPC · King of Industry | Speech Record
On July 24-25, 2026, the GIPC · King of Industries | 2026 Global Intellectual Property Ecology Conference, hosted by IPR Daily, was successfully held in Beijing. The conference adopts the methodology of "integration of law and business", focusing on three major industry topics: AI intellectual property services, biopharmaceutical IP value, and SEP global licensing. It brings together IP decision-makers from leading companies such as Xiaomi, Alibaba, Lenovo, CATL, and OPPO to jointly build an annual intellectual property industry event.
This column is a systematic compilation of the viewpoints of conference speakers, aiming to transform industry insights into reusable knowledge assets.
Yang Miao, Senior Director of Intellectual Property and Data Compliance at Jinshan Cloud, pointed out the structural contradiction faced by intellectual property protection in the AI era, where the system is measured by years and innovation is measured by weeks, with the title "Your Intellectual Property Strategy Should Replace in the AI Era". He proposed a new paradigm of "four layer armor+AI link" integrated protection: relying on technology locks to achieve instant protection, retaining core technical know-how with trade secrets, locking key innovation nodes with patents, and then combining copyright and open source layout to build an ecological barrier, relying on AI to achieve linkage and collaboration of the four layer protection system. Yang Miao emphasized that the core of breaking the deadlock lies in the triple thinking iteration: from protecting individual achievements to protecting innovation systems, from filling legal gaps to integrating research and development with intellectual property, from closed defense to establishing barriers in openness, and ultimately building a set of "visible, learnable, and unavoidable" AI era competitive moats.
The following is a transcript of the speech:
Since 2026, the large-scale implementation of Agent technology is profoundly reshaping the R&D pace of enterprises. The product iteration that used to be based on an annual cycle has now been compressed to units of weeks or even days. This kind of speed crushing directly impacts the existing intellectual property protection system - using traditional patent strategies is no longer effective in covering the actual needs of product protection. This is not the predicament of a single enterprise, but a structural contradiction that the entire industry is facing together.
Therefore, our company has comprehensively adjusted its intellectual property management strategy since the beginning of the year. Today, taking this opportunity, I would like to summarize my thoughts and practices during this period and exchange them with my peers. This strategy is not yet mature, I hope it can provide a useful breakthrough idea for reference.
Part.01
Why has the traditional IP strategy failed?
First, in the traditional Internet era, the complete cycle of a product from project initiation, grayscale test to official launch, to iterative update and even offline is usually calculated in years, and the intellectual property department has enough time to follow up, search, write and layout. But the AI era is completely different - the product concept proposed today may be launched in three weeks. R&D personnel no longer rely on lengthy documents to describe requirements, but directly use AI to generate runnable front-end pages, and only need to build databases and underlying links in the background. Testing is no longer done manually for three or four weeks, AI can complete it in half a day. The entire chain is so fast that the legal and IP teams have no time to react.
Secondly, when the code is generated with AI assistance, the design is illustrated by AI, and the text is polished by AI, the two most fundamental foundations of intellectual property, "originality" and "inventor," begin to loosen. In human-computer collaborative creation, the legal boundaries between human intellectual contributions and machine automatic outputs are not yet clear, but have become the norm in practice.
Thirdly, most mainstream AI products currently adopt a combination of open source and closed source strategies. The open source community not only accelerates technology diffusion, but also exposes core algorithms to the risk of being reverse engineered by competitors. The logic of competition has shifted from "closed defense" to "building barriers in openness", and the original moat is gradually being infiltrated.
Fourthly, in the past, companies were accustomed to using patents as the primary means of protection. But in the rapidly iterating environment of AI, the long cycle of patent authorization, the uncertainty of examination, and the risk of immediate disclosure make "relying solely on patents" impractical. It's like using only one shield to resist four arrows, losing sight of one.
Part.02
Breakthrough strategy: from individual combat to integrated defense line
In the face of the above difficulties, my core idea is to abandon dependence on a single means and build a fusion protection system of "four layers of armor+one link".
（1） The first layer of armor: technical lock, building the first line of technical defense
Technology lock refers to returning to the technology itself, using technical protection measures to safeguard intellectual achievements, and achieving the basic protection of 'not being able to steal'.
Ancient calligraphers and painters would bury exclusive identification marks in their works as proof of ownership. The intellectual property system itself is an imported product, relying on the legal system to achieve rights protection. But in the rapidly iterating environment of AI, we can return to simplicity and prioritize the release of the protective value of technological means. AI has significantly lowered the threshold for code writing, and the core value of research and development is shifting towards innovative thinking itself. Therefore, it is particularly crucial to retain ownership traces through technical means such as embedding technical tags, code watermarks, and keys.
The implementation of technology lock will face practical obstacles: R&D teams often believe that embedding tags and watermarks will increase their workload, and their willingness to cooperate is not strong. To promote implementation, it is necessary to push the system from top to bottom and collect feedback from bottom to top. Enterprises can incorporate technology protection into internal standards and set key deployment and traceability tagging as standard processes for source code writing. Different enterprises have different product forms, and can collaborate with technical teams to explore implementation methods that are suitable for themselves, such as digital embedding and access path control.
The core value of technology locks lies in striving for valuable time windows. Synchronize the deployment of technology protection actions during the research and development phase, and promptly synchronize them with the legal and intellectual property departments. Build the first wall for innovation before legal measures take effect.
（2） Second layer of armor: Dark armor, utilizing the value of trade secrets that are "strongest if not disclosed" effectively
The core logic of trade secrets is "protection without disclosure", and there is no statutory protection period. In the context of rapid iteration of AI products, once the technology is made public, it is easy for competitors to imitate or even replicate it with the help of AI analysis tools. Therefore, a large number of core technologies are more suitable for trade secret protection.
Currently, many enterprises have established systems related to trade secrets, but the management of trade secrets is often disconnected from the overall intellectual property strategy, and there is a lack of systematic coordination between the protection of technical secrets and the protection of product disclosure.
There is also a practical challenge in the implementation process: the assessment indicators of intellectual property departments are often guided by the number of patent applications, and the protection of trade secrets is an "invisible" work, which is difficult to intuitively reflect the results without infringement litigation. This makes the department subjectively more inclined to apply for patents, thereby abandoning the path of protecting trade secrets.
Therefore, enterprises need to establish a pre evaluation mechanism for technology protection methods, and assess in advance whether each technology is more suitable for patent disclosure protection or as a trade secret protection. The review subject can be entrusted to the technical committee, legal patent department, or management team according to the actual situation of the enterprise. At the same time, the management of trade secrets should be integrated into the entire R&D process, and complemented with public protection measures such as patents. Specifically, the core training dataset, tuning experience, prompt word engineering system, and intermediate version algorithms during the iteration process are all suitable for inclusion in the scope of trade secret protection.
（3） Third layer armor: Bright shield, heavily guarded at key patent nodes
Patents follow the logic of "open for monopoly" and are the most mature legal weapon in the field of intellectual property. But in the era of AI, patent strategy needs to shift from "comprehensive coverage" to precise layout.
A prominent challenge at present is that in the AI era, R&D personnel often master new technologies faster than intellectual property practitioners. Some R&D personnel use AI tools to directly read source code, automatically extract technological innovation points, and even generate technical disclosure documents ranging from ten to twenty pages. However, it is difficult to distinguish whether the disclosure documents generated by such AI are real innovative achievements or AI derived fictional solutions based on existing data, which poses a huge challenge to patent search and innovation point identification.
In this regard, patent layout can no longer pursue quantity, and must be heavily guarded at key nodes. The specific path is: first, the R&D leader verifies whether the innovation points belong to the core technology direction of the enterprise, and then uses AI tools for initial screening, combined with in-depth mining and retrieval by patent professionals, to finally complete the precise layout. We should focus on key anchor points such as core algorithm architecture and innovative interactive solutions that competitors cannot avoid, and make good use of acceleration channels such as pre-approval and PPH; For technical solutions that do not intend to apply for patents but need to prevent others from taking the lead, a defensive disclosure strategy should be adopted. Only in this way can the true value of patents as a legal "shield" be realized.
（4） The fourth layer of armor: weaving nets, copyright overlay, open source model, widely accumulating grain, building ecology
The combination of copyright and open source models can weave a three-dimensional protection network that covers both closed source and open source.
On the one hand, making good use of the natural advantages of copyright. Copyright automatically takes effect upon completion of the work, with low threshold and wide coverage. Enterprises should make ownership declarations for AI participation in generating content, complete copyright registration for core code, and protect the copyright of original annotation schemes, dataset structures, etc., to consolidate the rights foundation of closed source assets.
On the other hand, open source is an often underestimated competitive tool in the AI era. We can adopt a dual track strategy of "open sourcing the basic framework to attract the ecosystem, and closing the source of the core model and enterprise version to preserve the moat", using open sourcing to form industry technology standards and achieve ecological lock-in. At the same time, open source licenses should be carefully selected based on business objectives - for example, AGPL is suitable for defensive open source, and Apache 2.0 is suitable for ecological expansion. In certain scenarios, allowing more industry participants to use your technology is more helpful in building competitive barriers than simply prohibiting others from using it.
Copyright protection of closed source achievements, open source construction of ecological barriers, and synergy between the two can weave a protective network that combines offense and defense.
（5） Link: AI empowerment+deep integration
If the four layers of armor operate independently, there will still be protective gaps. AI tools are the link that connects the entire system.
At the tool level, AI can be implemented throughout the entire intellectual property process: automatically mining patent innovation points in R&D documents and code changes, conducting semantic patent searches, exponentially improving search efficiency, assisting in generating initial drafts of technical disclosure documents, continuously monitoring competitor trends, open source community dynamics, and preprint papers, achieving real-time warning of infringement risks; It can also monitor the code repository, identify the risk of trade secret leakage, and automatically draw a panoramic map of the enterprise's intellectual property assets.
But beyond tools, what is more important is the transformation of work modes. In the traditional mode, the legal and intellectual property teams only "fill in the gaps" after the research and development is completed, and the two systems are separated from each other. In the era of AI, it is necessary to prioritize intellectual property rights, embed them into the entire innovation cycle, and achieve integrated operations between R&D and intellectual property teams.
Specifically, an intellectual property liaison can be established within the R&D team to follow up on projects from the source; Conduct intellectual property evaluation simultaneously during the product requirement review and technical solution design phase; Organize joint review and dynamically adjust protection strategies after version iteration release; Simultaneously building a unified collaborative platform, connecting the information flow of R&D, legal, and management to ensure that protection strategies are updated in real-time with product evolution.
Part.03
Establish an asset protection decision map and select appropriate protection strategies
In the face of different types of innovation achievements, enterprises should not apply a unified template, but should establish clear decision-making logic, synthesize multi-dimensional judgments, and choose a single protection means or a combination of protection schemes.
Firstly, prioritize the technology lifecycle. Technologies with a lifecycle of less than 6 months should prioritize the combination of technology locks and trade secrets; For products with a lifecycle between 6 and 24 months, a fusion of technology locks, trade secrets, and patents is used for protection; For those with a lifecycle exceeding 24 months, patent protection is the main focus, supplemented by trade secrets.
Secondly, evaluate the competitive impact of technology disclosure. If it is made public, the core advantage will be lost, and trade secret protection should be chosen; If public disclosure would result in the loss of some advantages, it is advisable to adopt a combination strategy of patents and trade secrets; If public disclosure does not harm core values, copyright protection or open source models can be considered.
Thirdly, assess competitors' ability to bypass difficulties and meet the needs of ecological expansion. For core technologies that are easily bypassed by competitors, relying on patents to form blockades; For technologies with high difficulty in reverse replication, emphasis is placed on technical protection measures; If you need to build an industry ecosystem, you can adopt an open source strategy and choose different models such as core open source, value-added closed source, or layered open source.
After the implementation of this strategy, there may be a decrease in the number of patent applications, but this is not necessarily a bad thing. The reduction in quantity does not mean a reduction in the scope of work of the intellectual property department, but rather an expansion of the work scope and an improvement in comprehensive risk resistance ability. The shift from "quantity oriented" to "quality and effectiveness oriented" is itself the essence of this strategic adjustment.
Part.04
Conclusion
In the era of AI, intellectual property protection has never expired, but requires practitioners to become more sensitive and decisive. In the window period when new specialized legislation has not yet been implemented, we do not need to wait for new rules, but should make good use of existing institutional tools and actively build an integrated protection system.
Our ultimate goal is to create a set of competitive barriers that are visible, difficult to learn, and unavoidable. To achieve this, a triple iteration of thinking is needed: upgrading from protecting a single achievement to protecting the entire innovation system; Engage in post legal position filling and upgrade to integrated R&D and legal collaborative operations; Upgrade from closed defense to building barriers in openness.
A single moat cannot withstand the impact of rapid iteration in the AI era. Only a fusion defense line that intertwines technology locks, trade secrets, patents, copyrights, and open source can truly safeguard the core innovation advantages of enterprises.
Thank you all!
69F, Tower A, Ping An Finance Center, No. 5033 Yitian Road, Futian District, Shenzhen, Guangdong Province 518000, P.R. China
+86-755-22984592/+86-755-82094518
Patent & Others: zysw@zyip.com 
Patent
Trademark
Litigation
Copyright
Project
Other
+86-755-22984592
zysw@zyip.com