The integration of legal large language models into legal services has sparked ethical controversies—how should lawyers address the question of responsibility attribution in the algorithmic age?
Introduction: When algorithms begin to participate in legal judgments, where do the boundaries of lawyers lie?
On August 29, 2026, the 10th Emerging Legal Services Industry Development Forum was held at the Koguan School of Law, Shanghai Jiao Tong University, where over 200 tech pioneers and legal professionals gathered to discuss "Justice in the Age of Algorithms." The "Emerging Legal Services Industry Development Report (2026)" released at the forum revealed that the global legal technology market is projected to reach $36.72 billion, with legal large language models evolving from auxiliary tools into "digital collaborators" deeply involved in decision-making.
As a lawyer, what concerns me is not just the dazzling technology, but the legal issues surging beneath it: when AI assists in contract review, case prediction, and even sentencing recommendations, who bears the responsibility if errors occur? When client data flows through the cloud, how can confidentiality obligations be upheld? When law firms embrace AI, how can they prevent "technological dependence" from sliding into "professional failure"? This article will, from the perspective of legal practice, break down the core risks that emerge when algorithms are embedded into legal services, and offer strategies for addressing them.
AI's involvement in legal decision-making makes the assignment of responsibility the biggest "hidden reef."
At the forum, multiple experts noted that AI can already handle tasks such as information retrieval and document generation, but the final judgment still requires lawyers to review. This raises a key question: if AI-generated legal advice is flawed and causes client losses, who bears the responsibility?
Under current law, Article 1165 of the Tort Liability Book of the Civil Code establishes the principle of fault-based liability. If a lawyer fails to exercise due diligence in reviewing AI output, they remain liable for professional negligence. Contracts between law firms and AI suppliers may involve product liability or defective services, but clients typically hold only the lawyer accountable—after all, the engagement agreement is signed under the law firm's name.
In practice, I have seen a law firm use an AI contract review tool that missed a critical liquidated damages clause, causing the client to suffer losses after signing based on that contract. The lawyer defended by citing a "tool defect," but the court did not accept this, ruling that the lawyer is responsible for the final text. This serves as a warning: AI is the "co-pilot," but the steering wheel remains in the lawyer's hands—no technology can replace professional judgment. If a law firm adopts AI, it must establish a manual review mechanism and clearly define the boundaries of AI use in internal protocols; otherwise, any mishap becomes a professional liability risk.
Client data moving to the cloud presents new challenges for confidentiality obligations.
Legal AI products require extensive data training, and when used by law firms, they often involve client trade secrets and personal privacy. The intelligent agent products showcased at the forum emphasize "confidential-state inference and spatial isolation," but in reality, many small and medium-sized law firms directly use public cloud AI tools to process case files, leaving the flow of data shrouded in mystery.
Article 38 of the Lawyers Law explicitly imposes a duty of confidentiality on lawyers, while Article 28 of the Personal Information Protection Law sets a higher threshold for the processing of sensitive personal information. If client data is uploaded to third-party AI platforms without desensitization, once a leak occurs, the law firm not only faces administrative penalties but may also be subject to claims for compensation from clients. Last year, a law firm in a certain region was summoned for talks by the cyberspace administration authorities due to using an overseas AI tool to translate confidential documents—a profound lesson.
My recommendation is that law firms should prioritize legal tech products that have passed Class III Cybersecurity Protection Certification and sign data processing agreements that clearly specify data storage locations, purposes of use, and deletion mechanisms. For cases involving confidential information, it is better to handle them manually than to rush to the cloud for speed. Individual lawyers should also develop a "minimization" habit—uploading only necessary information and avoiding exposing entire case files without protection.
Algorithmic bias may "skew" the referee's reasoning.
At the forum, Professor Peng Chengxin from Shanghai Jiao Tong University emphasized "balancing the instrumental rationality of algorithms with the value rationality of law," directly pointing to the risks of algorithmic bias. If AI training data is derived from historical court judgments, it may entrench certain biases, such as imposing harsher sentences on specific groups. If lawyers rely on AI for similar-case retrieval, they may be "led by the nose," overlooking the differences in individual cases.
Article 6 of the Judges Law requires judges to adjudicate independently in accordance with the law, but if AI becomes involved in sentencing recommendations, it may create "hidden injustice." Although AI is currently mostly used as an auxiliary tool, lawyers must remain vigilant: Are the retrieved similar cases representative? Is the strategy recommended by AI suitable for the case at hand? I once handled a labor dispute where the AI search results leaned toward supporting the company, because the database mostly contained cases won by employers. Only after manual review did we find the key precedent in favor of the employee.
Lawyers should treat AI as a "junior assistant," not an "authoritative mentor." After conducting similar-case searches, manual screening and comparison of case causes and factual details are required, followed by judgment based on local judicial policies. Especially in sensitive areas such as personality rights and labor disputes, independent critical thinking must be maintained to avoid "technological superstition."
Law Firm Digital Transformation: How to Balance Efficiency and Compliance?
In the "Trend List" published on the forum, many award-winning products focus on contract review and case management. However, for law firms, introducing AI is not just about purchasing software—it is an organizational transformation. Liu Qiming from Zhihe proposed that future law firms should build an architecture of "legal intelligent agents + super lawyer teams," but during the transition, a "two-tier" problem often emerges—where the technology department pushes AI, lawyers find it troublesome and refuse to use it, or they use it without procedural constraints.
From a compliance perspective, law firms should establish an "AI Usage Management Measures" that clearly defines the scope of application, approval procedures, and data desensitization requirements. For example, which documents can use AI for initial drafts? Which must be written manually? Should AI-generated content be labeled? These details can reference Article 9 of the "Interim Measures for the Management of Generative AI Services," which requires AI service providers to bear responsibility for content security, but law firms, as users, also need to establish their own review mechanisms.
I suggest that law firms establish an "AI Ethics Committee," composed of senior lawyers and technical experts, to regularly assess tool risks. For clients, firms should proactively disclose the scope of AI usage and obtain informed consent to avoid future disputes. After all, technology serves people; if the transition costs us trust, it would be a loss outweighing the gain.
Conclusion: Amid the waves of technology, the anchor of a lawyer's value lies in the human element.
In the age of algorithms, the question of justice must ultimately be answered by humans. Legal AI can enhance efficiency, but it cannot replace a lawyer's empathy, judgment, and sense of responsibility. For individuals and businesses, when choosing a law firm, it is advisable to inquire about its AI usage policies and prioritize institutions with clear compliance mechanisms. For lawyers, while embracing technology, it is even more crucial to cultivate their core competencies—understanding statutes, analyzing evidence, and communicating persuasively—these are the irreplaceable moats.
Guangdong Zhiming Law Firm has been consistently focusing on the application of legal technology in recent years, exploring AI-assisted case handling while adhering to the bottom line of manual review. We are willing to discuss with our peers how to strike a balance between efficiency and justice, ensuring that technology truly serves the rule of law. If you have any questions regarding legal risks related to AI, please feel free to consult us, and we will provide professional advice.
