As large language models enter China’s legal profession, which lawyers will lose out?

A major legal database affiliated with Peking University has launched a large language model tool it says can accurately retrieve statutes and automatically generate contracts, sending ripples through the legal profession in China.
AI-powered legal tools can draft convincing documents in seconds, yet without rigorous oversight they are equally adept at inventing statutes and fabricating precedents – a liability that has kept them on the sidelines in medicine and law.
However, lawyers cannot afford to ignore the technology; the real question is not whether to use it, but how to harness its speed without sacrificing trust.

Enter Chinalawinfo PKULaw, Peking University’s flagship legal database, which has rolled out a standard Model Context Protocol (MCP) interface – developed by Anthropic in 2024 – that plugs into any large language model like a flash drive, instantly equipping it with authoritative legal retrieval.
On its WeChat account, Chinalawinfo PKULaw said the new service – with backing from a vast repository of regulations, court rulings, academic analyses and case records – allowed users to search, verify currency, draft contracts and collate similar cases, all while delivering outputs that were not only accurate but also traceable to their source.
In short, the service transforms generative AI from a black box of potential hallucinations into a transparent, verifiable research partner, according to the developers.
But will it replace human lawyers?
Zhang Xian, deputy general manager of Chinalawinfo PKULaw, said the service was positioned squarely as an assistant, not a replacement.
“The first groups to feel the impact [are] likely to be legal assistants, junior lawyers, legal specialists, legal researchers and those responsible for search, organisation, comparison and preliminary drafting,” Zhang said last week.
“However, while AI improves the efficiency of basic information processing, it cannot replace legal professionals’ judgment, accountability or client communication. The more important direction going forward will be human-AI collaboration, allowing legal experts to devote more energy to complex decision-making and high-value services.”
In large enterprises, where highly repetitive and scalable tasks are common, AI appears to have an even clearer role.
In one official case study, in-house legal staff needed no programming skills – they simply “translated” their professional experience and judgment into instructions for the AI. The system, which was integrated into office software, could then receive complaints and contract materials, analyse the claims and legal basis, and stamp and submit documents.
Zhang said the Chinalawinfo PKULaw legal intelligent search MCP service had already partnered with Tencent and Alibaba. Alibaba owns the South China Morning Post.
Yet, a corporate legal officer in Beijing said that such software was not likely to entirely replace senior staff any time soon, because the use of these tools and the input of instructions still required professional judgment.
“The skill threshold for fine-tuning AI assistance remains high; when handling specific cases, legal reasoning must be deconstructed into a sequence of operations executable by the AI. Without legal experience, a user who asks for a rough document will only receive a very basic draft in return,” she said.
Nevertheless, junior employees who previously handled such paperwork may well be displaced.
“The industry has become unwelcoming to inexperienced newcomers, prompting many to switch careers,” the Beijing-based lawyer added. Her alma mater, a prestigious Beijing law school, saw record numbers of graduates taking civil service exams last year.
A sole practitioner in the northwestern city of Xian echoed the sentiment, admitting that she had previously dared not use AI for legal research, as off-the-shelf models often fabricated statutes when responding to legal queries. Instead, she always turned to professional databases for verification.
“If a reliable tool could handle these paperwork tasks, it would dramatically boost efficiency,” she said. But, based on her own experience, she found that in many ordinary civil cases, evidence alone often carried the day and there was little need to delve into legal doctrines – making human coordination even more critical in those circumstances.

Zhang Xian said Chinalawinfo PKULaw’s MCP service could overcome faked responses because of its evidence-based technology stack – a closed loop, from understanding the question to dynamic updating, that curbed hallucinations – and its real-time sync with the PKULaw database that fed new laws and cases into the AI.
“AI technology is still developing rapidly. Our view is that AI will not replace the legal profession, but it will certainly change the way legal services are produced,” Zhang said.
“Search, organisation, drafting and verification will become increasingly intelligent. However, complex judgment, client communication, strategic decision-making, professional ethics and responsibility will remain the core value of legal professionals.”