Behind the millions of financing for legal AI: Three major legal risks and compliance boundaries for lawyers using AI to handle cases
Lead
In September 2026, the legal AI agent product "Yongjian AI" announced the completion of its Series A financing, just two months after its launch. The investor is not a traditional VC, but a commercial IoT company, Shangmi Technology, and a level protection evaluation agency certified by the Ministry of Public Security, Golden Shield Testing. This financing news has attracted attention in the legal technology industry, but what is more worthy of scrutiny by the legal community is the legal issues involved behind the product: when AI starts to "work" for lawyers - reviewing contracts, conducting due diligence, writing documents, and searching for cases - how to define the professional responsibilities of lawyers? Where is the security boundary of customer data? Who is responsible for the consequences of AI generated content errors?
As lawyers deeply involved in legal practice, we may start from this financing incident and dissect the legal risks and compliance points that cannot be avoided after legal AI enters the lawyer's workflow.
1、 AI Illusion and Professional Responsibility: Lawyer's Signature Must Be 'Blamed'
The most fatal flaw of legal AI is "illusion" - generating seemingly reasonable but actually fictional precedents, laws, or legal viewpoints. Yongjian AI emphasized its retrieval ability of "traceable source" in its publicity, covering many jurisdictions such as Chinese Mainland, Hong Kong, Singapore, and so on, and its articles and cases all point to the original source. This precisely indicates that the industry has realized that the issue of illusions is the first life and death hurdle for the commercialization of legal AI.
From the perspective of legal practice, the core of the problem lies in Article 49 of the Lawyer Law, which stipulates that "if a lawyer engages in illegal practice or causes losses to a party due to fault, the law firm where the lawyer works shall bear the liability for compensation. If a lawyer submits a legal opinion generated by AI containing fictional precedents to a client or court, and the client makes a business decision based on the opinion and suffers losses, the lawyer and law firm will face breach of contract or infringement compensation. More seriously, if false precedents are submitted to the court, it may violate Article 114 of the Civil Procedure Law on compulsory measures for obstructing civil litigation, and face fines, detention, and even criminal liability.
In judicial practice, there are precedents of courts punishing lawyers who use AI to generate false cases. In 2025, a court in a certain place discovered during a contract dispute trial that the "similar case judgment" submitted by the proxy lawyer was fabricated by AI, and ultimately gave the lawyer a warning. This signal is very clear: AI is a tool, but it is the lawyer who signs it, and the responsibility cannot be shifted.
Therefore, lawyers must establish a mandatory process of "manual review" when using any legal AI tool. According to Article 32 of the Measures for the Administration of Lawyer Practice, lawyers shall "make independent judgments on the facts of the case and the application of the law" and shall not completely transfer the core legal judgment power to algorithms. AI can assist in retrieving, organizing materials, and generating initial drafts, but the final verification obligation, judgment obligation, and signature responsibility always lie with the lawyer themselves.
2、 Customer data security and confidentiality obligations: Can case materials be uploaded to the cloud?
The second sensitive point of legal AI is data security. Case materials, contract texts, and due diligence materials often contain clients' trade secrets, personal privacy, and even state secrets. Yongjian AI adopts a desktop agent form, and data is left on the local machine by default. Cloud calls are only made through the official API of the model selected by the user. The platform does not save task context and supports local model solutions. These designs are essentially in response to the lawyer's confidentiality obligation stipulated in Article 38 of the Lawyer Law.
Article 38 of the Lawyer Law clearly states: "Lawyers shall keep confidential any state secrets or trade secrets they become aware of during their professional activities, and shall not disclose the privacy of their clients. ”Meanwhile, Articles 1032 to 1039 of the Civil Code establish complete rules for privacy and personal information protection. If a lawyer uploads case materials containing the privacy of the parties to an AI platform that has not passed the security assessment, once data leakage occurs, they may not only face civil compensation, but also be warned, fined, or even suspended from practice by judicial administrative authorities in accordance with the "Punishment Measures for Illegal Acts of Lawyers and Law Firms".
In 2025, a law firm was fined millions of yuan by a client for using overseas AI tools to handle a trade secret case, resulting in the leakage of their technical information. The firm's partners were also disciplined by the industry. This case serves as a warning: when choosing legal AI tools, it is necessary to review key compliance indicators such as data storage location, transmission encryption method, context retention, and whether they are used for model training.
From a practical perspective, lawyers and law firms should establish an AI tool access system: case materials involving state secrets are strictly prohibited from using any cloud based AI; for cases involving trade secrets and personal privacy, priority should be given to local deployment or desensitization solutions; Sign data processing agreements with AI service providers to clarify their confidentiality obligations and breach of contract responsibilities.
3、 Copyright and evidential validity of AI generated documents: Who possesses the 'skills'?
Yongjian AI regards "skills" as the core of its products, allowing lawyers to deposit their personal work methods into reusable digital assets and promote the skills market. This raises a cutting-edge legal question: Who owns the intellectual property rights of the "skills" formed by the collaboration between lawyers and AI, including case thinking, risk assessment standards, and document templates?
According to Article 3 of the Copyright Law, works must have originality. If "skills" reflect the personalized selection, arrangement, and judgment of lawyers, they may constitute works protected by copyright law. But if the skills are mainly generated automatically by AI based on general rules, they may not be protected due to a lack of human original contributions. At present, the judicial practice has not yet formed a unified standard, but the "AI generated content copyright case" heard by the Beijing Internet Court in 2025 has confirmed that AI auxiliary achievements reflecting human intelligence investment can constitute works.
Another key issue is the evidential validity of AI generated documents. According to Article 90 of the Several Provisions of the Supreme People's Court on Evidence in Civil Litigation, the authenticity, legality, and relevance of electronic data submitted by the parties shall be reviewed. If AI generated case retrieval reports and legal opinions are submitted as evidence, the other party may question the reliability and completeness of their sources. Lawyers should retain the original data, retrieval records, and manually modified traces generated by AI to form a complete chain of evidence to address cross examination challenges.
4、 Industrial capital entry and law firm compliance management: legal review in AI procurement
Shangmi Technology and Jindun Detection's investment in AI indicates that legal AI is evolving from a pure software tool to an ecosystem of "hardware+security+scenarios". Shangmi covers commercial IoT terminals in over 200 countries and regions worldwide, while Jindun Testing has long been engaged in government and enterprise security and security evaluation. This industrial capital structure itself sends a signal that the competition of legal AI has entered the deep waters of data security, compliant storage, and scenario implementation.
For law firms, purchasing AI tools is no longer just a decision for the IT department, but a compliance action that requires legal review. Law firms should conduct data security risk assessments on AI service providers in accordance with Article 27 of the Data Security Law and Article 51 of the Personal Information Protection Law, clarifying whether they have passed level protection assessments, whether they have compliance pathways for data export, and whether they have established emergency response mechanisms for data breaches.
At the same time, the contract between law firms and AI service providers should clearly stipulate that service providers shall not use law firm data for model training; Data storage period and destruction method; The obligation to notify and compensate in the event of a data breach; The mechanism for recovering losses incurred by law firms due to AI generated content errors.
Conclusion: AI is a tool, compliance is the bottom line
The shift of legal AI from "Q&A" to "work" can indeed improve the efficiency of lawyers' work, but efficiency cannot be achieved at the expense of professional safety and client trust. Lawyers must adhere to three bottom lines when using AI tools: AI generated content must be manually verified, customer data must be secure and controllable, and professional responsibility must be borne by themselves.
Guangdong Zhiming Law Firm has long been concerned with compliance issues in the field of legal technology, and can provide professional legal services such as AI tool procurement compliance review, data security system construction, and AI generated content infringement for law firms and lawyers. In the current era of rapid technological iteration, only by having professional lawyers help you clarify boundaries can AI truly become a help rather than a hidden danger.
