iFlytek Releases iFlytek Xiaofa Super Legal Assistant, Integrating Software and Hardware to Solve the Trustworthiness Challenge of Legal AI

Special Topic: iFlytek Global 1024 Developer Festival — New Crowdtesting Promotional Copy Added — New Crowdtesting Promotional Copy Added end — App-Exclusive Traffic Guidance Added

On October 22, the Trusted Legal AI Practice Forum and iFlytek Xiaofa New Product Launch was held at the China National Convention Center in Beijing. This event was one of the series of activities of the iFlytek 1024 Developer Festival. Representatives from the legal academic community, legal affairs departments of central state-owned enterprises, leading law firms, and the technology industry gathered to discuss the trusted practice and industrial co-construction of legal technology amid the wave of vertical large models.

At the meeting, iFlytek officially released the iFlytek Xiaofa Super Legal Assistant. This is a software-hardware integrated solution composed of the iFlytek Xiaofa Super Legal Assistant, the iFlytek Xiaofa Recording Card, and the iFlytek Xiaofa Smart Box. Relying on the Spark Legal Large Model and the device-cloud collaborative architecture, it is aimed at corporate legal departments and lawyer teams, focusing on solving the industry pain point that legal AI "can be used but is not dared to be used."

Zhang Youguo, Senior Vice President of iFlytek, said in his speech that domestic legal technology has completed the stage of "from nothing to something" and is now entering the critical moment of "from something to excellence." The legal industry has very little room for error; once there is a mistake in citing legal provisions or calculating time limits, it may bring irreversible consequences. Data security for sensitive materials such as case files and trade secrets is also a real challenge the industry must face. "The legal industry cannot tolerate 'roughly correct.' We hope that every conclusion can withstand verification and that every case file remains in the hands of legal professionals."

Zhang Youguo introduced that iFlytek has been deeply engaged in the judicial field for many years and has always insisted that technology must not only understand the law and business, but also firmly uphold the security bottom line. The new product does not rely on verbal promises to build trust, but uses mechanisms to make conclusions verifiable and case files controllable, while leaving the final legal judgment to legal practitioners. He said that legal AI cannot be polished behind closed doors in a laboratory; it must take root in frontline practice and continue to iterate. iFlytek is willing to be an industry companion and work with political and legal institutions, universities, and law firms to build a trusted and win-win smart rule-of-law ecosystem.

Shi Jianzhong, Dean of the Data Law School (Data Rule of Law Research Institute) of China University of Political Science and Law, shared his views on the reshaping of responsibilities and compliance boundaries of intelligent agents empowering rule-of-law technology products. He pointed out that intelligent agent technology is driving a leap in the capabilities of rule-of-law technology products. Products not only have proactive execution capabilities, but also use multimodality and embodied intelligence to break through the physical boundaries of rule-of-law services. At the same time, technological iteration also brings many complex risk challenges. Rule-of-law technology products should set industry benchmarks, adhere to the principles of stability, reliability, and traceability, and fully demonstrate and strengthen the authority of the rule of law.

Shi Jianzhong emphasized that the core essence of rule-of-law technology products lies in improving the efficiency of achieving fairness and justice. Product design should embed due process within it, mandatorily establish a full-process mechanism for human intervention in deconstruction, review, and judgment, conduct decomposition and analysis of facts, evidence, and legal provisions, ensure that reasoning paths are traceable, firmly uphold the final decision-making power of humans, safeguard the bottom line of justice through rigorous procedure, and fully unleash human wisdom through machine computing power.

Yong Wenyuan pointed out that more and more legal practitioners have begun to use AI tools; however, "knowing how to use" does not equal "daring to use," and "being usable" does not equal "being trustworthy." Industry research shows that legal AI faces four underlying gaps: first, data security risks create concerns about use; second, model hallucinations create a credibility crisis for conclusions; third, verification costs are too high, offsetting the efficiency gains brought by AI; fourth, tool fragmentation splits the complete case-handling process.

He believes that global competition in legal technology has shifted from single-point tools to end-to-end complete workflows. Domestic legal AI cannot remain at the level of competition in single-point capabilities such as document generation and legal provision retrieval; instead, whether a complete case can be handled from beginning to end should be the core measurement standard. Products should be embedded in the complete chain of legal professionals from case intake to case closure, build trusted mechanisms for security, traceability, verification, and full-process integration, and truly turn AI from "a tool that adds flowers to brocade" into a business foundation that can be relied upon in practice.

Software-Hardware Integration, Device-Cloud Collaboration: iFlytek Xiaofa · Super Legal Assistant Officially Released

At the event, Zhan Longyun, Deputy General Manager of Spark Xiaofa (Shanghai) Technology Co., Ltd., officially released the iFlytek Xiaofa Super Legal Assistant. The starting point of this product is very clear: case files always remain in the hands of lawyers, and AI, within secure boundaries, helps lawyers handle cases faster and more accurately.

iFlytek entered judicial scenarios starting from courtroom speech recognition. Currently, more than 20,000 courtrooms nationwide use iFlytek speech systems to assist in producing court hearing transcripts. Since then, it has extended large model capabilities to judges' and prosecutors' case file review and document work.

On this basis, iFlytek Xiaofa serves lawyers and corporate legal departments. With the Spark Legal Large Model trained on domestically autonomous and controllable computing power as the foundation, it builds core capabilities around five directions.

More secure: The system automatically schedules between device and cloud. Confidential case files are processed locally and desensitized; after similar-case retrieval and cross-verification are completed in the cloud, they are restored locally with one click. Paired with the iFlytek Xiaofa Smart Box, it can achieve full-process intranet offline inference, so original case files never leave the intranet.

More trustworthy: It achieves character-level bidirectional traceability, and every legal viewpoint can be located in the original text of the case file. Legal provisions and similar cases are dynamically updated daily, repealed regulations are automatically warned, the reasoning chain is fully retained, and conclusions can be reviewed and cross-examined.

More professional: It goes beyond text organization and, from the client's standpoint, completes evidence review, extraction of disputed issues, and courtroom risk alerts to serve litigation analysis and judgment.

More comprehensive: It covers the entire process from case intake and file creation to case closure and archiving. The iFlytek Xiaofa Recording Card can record in real time, transcribe field investigation content, and synchronize it to the case file. Open interfaces are compatible with lawyers' existing office systems, eliminating the need to repeatedly import and export between multiple software programs.

More understanding of you: three-layer memory across cases, individuals, and organizations, supporting resumption of work across cycles at breakpoints, adapting to personal writing styles, and consolidating law firm templates, cases, and risk control guidelines into team knowledge assets.

This assistant supports both cloud and local Smart Box deployment methods, can handle ultra-long case files of 30,000 pages, and supports desensitization and restoration of 17 types of sensitive information. At present, it has been trialed in some law firms. Based on usage statistics from nearly 200 frontline lawyers and legal affairs personnel, overall case-handling efficiency has improved by more than 45%.

Roundtable dialogue: jointly exploring the path of co-building legal technology in the era of vertical large models

In a roundtable dialogue themed "Trusted Practices and Industrial Co-building of Legal AI," Zhu Jing, Vice President of Legal Affairs and General Manager of the International Development Department of iFlytek, served as moderator. Yu Mang, General Manager of the Legal and Regulatory Affairs Department of China Mobile Communications Group Co., Ltd. and Vice President of the China Law Society's Internet and Information Law Research Institute; Jin Hao, Vice President and General Counsel of ChangXin Technology Group Co., Ltd.; Zhang Xiaowei, Director, Senior Partner, Director of the National Management Committee, and Party Committee Secretary of Beijing Weiheng Law Firm; Sheng Bin, member of the Executive Committee of the Management Headquarters, Deputy Party Committee Secretary, and Senior Partner of Shanghai AllBright Law Offices; and Li Hongjiang, Deputy Party Committee Secretary, Executive Partner, and Director of the Intellectual Property Committee of Beijing Guantao Law Firm, participated in the discussion.

Yu Mang stated that AI is suitable for undertaking standardized work such as material organization and data retrieval, while core legal judgments must be made by humans. As enterprises accelerate their "going global" efforts, domestic legal AI should also fill gaps in capabilities such as ascertaining foreign law and supporting cross-border disputes, relying on a self-controllable technology foundation and domestic rule-of-law practice data to form advantages that distinguish it from overseas products.

Jin Hao believes that the boundary of human-machine rights and responsibilities should be dynamically adjusted as technological capabilities evolve: high-frequency, standardized, and reviewable work can be handed over to AI, while matters involving value judgments and major commercial decisions must be checked by humans. ChangXin encourages its legal team to discover problems and provide feedback on needs during real use, using practice to drive product iteration.

Zhang Xiaowei introduced that Weiheng sets red lines for lawyers' AI use through regular risk warnings, strictly prohibits the irregular use of overseas AI tools, and strictly prevents the leakage of client cases. He believes that independent research and development of legal AI by large law firms requires large investment and is highly difficult, and collaborative co-building with professional technology vendors is a more efficient path. iFlytek Xiaofa Smart Box directly addresses law firms' most concerning pain point: data confidentiality.

Overseas mature legal AI models are built on the common law system and are difficult to directly transplant to China's statutory law environment. Sheng Bin's suggestion is that law firms can combine their own business characteristics to explore vertical legal agents in niche tracks such as capital market due diligence and cross-border legal services. In recent years, AllBright has steadily advanced AI localized deployment and digital-intelligent construction, adhering to the principles of independent controllability, security, and trustworthiness. At present, AllBright and iFlytek have reached a strategic cooperation. Relying on iFlytek's legal large model technology foundation, they are exploring intelligent application solutions around litigation, non-litigation, and foreign-related legal service scenarios, and building vertical legal intelligent tools that fit practical scenarios.

Li Hongjiang introduced that Guantao has issued AI use management regulations, requiring desensitization of materials before use, item-by-item manual verification of output content, and clarifying that responsibility for use is borne by the user. He believes that vertical legal AI can narrow the technology gap between large firms and small and medium-sized firms and lower the threshold for legal services.

The guests unanimously agreed that legal AI cannot simply pursue technical indicators; it should establish humans as the ultimate responsible subject and hold the two bottom lines of data security and trustworthy results. For trusted legal AI to mature, technology vendors need to go deep into frontline case-handling scenarios, work together with corporate legal affairs and law firms in practice, and jointly build an open, collaborative, stable, and controllable smart rule-of-law ecosystem. Statement: This is an exclusive article of Sina.com. Reproduction without authorization is prohibited. tech_web_article_v2015_block_bottom Non-targeted 300*250 button 17/09 wenjing begin

Non-targeted 300*250 button end