Futures
Access hundreds of perpetual contracts
TradFi
Gold
One platform for global traditional assets
Options
Hot
Trade European-style vanilla options
Unified Account
Maximize your capital efficiency
Demo Trading
Introduction to Futures Trading
Learn the basics of futures trading
Futures Events
Join events to earn rewards
Demo Trading
Use virtual funds to practice risk-free trading
Launch
CandyDrop
Collect candies to earn airdrops
Launchpool
Quick staking, earn potential new tokens
HODLer Airdrop
Hold GT and get massive airdrops for free
Pre-IPOs
Unlock full access to global stock IPOs
Alpha Points
Trade on-chain assets and earn airdrops
Futures Points
Earn futures points and claim airdrop rewards
What are the risks of using intelligent agents like "little lobsters" to write patent application documents, and how can they be avoided?
The China National Intellectual Property Administration issued a risk alert yesterday (the 1st). It said that smart agent tools such as OpenClaw (“Lobster”) have been exposed for having weak default security configurations, which can easily lead to serious security risks. When using such smart agents to write patent application documents, there may also be multiple hidden hazards.
Previously, when companies applied for an invention patent, it often took a long time and cost a lot. Even just drafting a patent application document required considerable effort. Now that artificial intelligence is available, a set of claims can be generated in a very short time, with costs of only a few hundred yuan. But behind the “AI acceleration” there are also many risks.
At present, regulators and technology companies in multiple countries have already rolled out usage guides and rules regarding “OpenClaw.” What issues need to be considered when using AI smart agents to draft patent application documents?
The China National Intellectual Property Administration recently issued a risk alert. Using smart agents such as “Lobster” to draft patent application documents may trigger multiple risks. The first is the risk of “technical information leakage.”
A person in charge from the Patent Examination Business Administration Department of the China National Intellectual Property Administration: Regarding the risk of technical information leakage, smart agents such as OpenClaw have issues such as excessively high permissions, security vulnerabilities, and plugin poisoning. If they are used to write application documents, it is easy to cause core information such as the technical disclosure letter to leak. Once the technical solution you intend to apply for leaks out, the patent application’s technical scheme may fail to be authorized due to loss of novelty, and it may even be filed first by someone else, causing major losses to the applicant. The agency would also need to bear liability for breach and compensation.
Zhao Zhanling, a special research fellow at the Intellectual Property Research Center of China University of Political Science and Law, explained that, according to the relevant provisions of China’s Patent Law, the inventions and utility models for which patent rights are granted must be novel. Once core information such as the technical disclosure letter is leaked through a smart agent that has security risks, it constitutes an act of disclosure. Simply put, once it is “disclosed,” it is considered “not novel,” so it cannot be authorized.
The China National Intellectual Property Administration said that, in the field of patent applications, the issue of AI hallucinations can also lead to risks of “substantive defects.”
A person in charge from the Patent Examination Business Administration Department of the China National Intellectual Property Administration: When using this kind of smart agent to draft application documents, “AI hallucinations” may occur, causing the application documents to be seemingly right but actually wrong, with logical contradictions in the content, and unclear expression of technical features, thereby making it impossible to obtain protection.
The China National Intellectual Property Administration also pointed out that patent application submissions formed by smart agents through generating out of thin air, inventing at will, or combining content, constitute an untrustworthy behavior of patent application that violates the principle of good faith, and there is a risk of “untrustworthy patent application.”
A person in charge from the Patent Examination Business Administration Department of the China National Intellectual Property Administration: If the approach of “untrustworthy patent applications” reaches a certain number, the applicant will face administrative penalties such as warnings and fines. The patent agencies and patent agents will face administrative penalties such as revocation of practice licenses and revocation of authorization qualifications for patent agencies. In serious cases, they will be listed in the list of serious violations of law and dishonesty.
After reviewing publicly reported coverage, journalists found that in the field of patent applications, some localities have used artificial intelligence technology to launch smart agent or large-model service platforms, providing enterprises or individuals with functions such as online consultants, intelligent error correction, and document retrieval during the patent application process—essentially a “smart butler” for patent applications.
For example, in a certain locality in Jiangsu Province, an intellectual property public service AI smart agent was launched earlier. Its core functions include automatically generating drafts of patent technical disclosure letters and optimizing them according to standards, intelligent search, and patent comparison and analysis. It can help small and medium-sized enterprises “reduce costs and burdens,” and substantially compress the cost of professional services. Another locality in Zhejiang Province launched an AI intellectual property generation platform. As long as the company uploads product images, after a few minutes a clear-structure patent “claims” document can be generated.
How should the traditional intellectual property system adapt to the development of AI technology, and what institutional innovations are needed? Zhao Zhanling said that to adapt the traditional intellectual property system to AI technology development, the first task is to clarify in the Patent Law the standard of substantial contribution, define the extent to which natural persons exercise intellectual control over the conception, verification, and expression of technical solutions, exclude authorization scope for results generated purely by artificial intelligence, and protect the incentive logic of the patent system.
Zhao Zhanling, a special research fellow at the Intellectual Property Research Center of China University of Political Science and Law: The “Patent Examination Guidelines” need to add special examination standards for AI-generated documents. Examiners have the right to require applicants to disclose the process through which the technical solution was formed, to identify whether it is a trustworthy patent application. On the other hand, if there are technical methods that can determine whether the application document was generated by AI, it would be even better—so as to determine where the applicant’s substantial contribution lies.
As AI technology empowers industries across the board, how should small and medium-sized enterprises and individual inventors use AI tools safely in the future? Zhao Zhanling suggested that when inventors use AI tools, they should strictly limit them to auxiliary steps such as information retrieval, language polishing, and format organization, while the core technical scheme conception, claims design, and legal risk assessment should be completed personally by the inventor or reviewed by professional patent agents.
Zhao Zhanling, a special research fellow at the Intellectual Property Research Center of China University of Political Science and Law: It is necessary to establish a complete usage traceability mechanism—save records of every interaction with the AI, versions that were manually edited, and the final reviewed and confirmed documents. By maintaining these records, it can be demonstrated that the core technical documents were written or confirmed by humans. The most critical point, however, is that when choosing a patent agency, applicants should proactively ask whether they will use AI, and even require a clear commitment to human drafting, with the AI used only as an aid. In particular, have the relevant commitments written into the entrustment contract to mitigate risks. If the patent agency violates the agreement, the patent agency can be held responsible.
A wealth of information and precise analysis—everything is right here in the Sina Finance APP