In Depth: Legal Professionals See a Case for AI, but It’s Not Ironclad

Artificial intelligence (AI) is rapidly transforming the legal profession, moving from theoretical conversations to everyday applications across law firms, corporate legal departments, and courts [para. 1]. A recent LexisNexis survey of 404 Chinese legal professionals revealed that 88% already use AI in their work, a steep increase from the previous year. The figure reaches 93% among lawyers and trainees and 86% among compliance and risk officers. Common uses include drafting emails, memos, contracts, and summarizing or condensing legal documents and case files [para. 2][para. 3].

Judicial systems are experimenting as well. In January 2023, shortly after ChatGPT's release, a Colombian judge admitted using it in a ruling. By late 2024, China's Supreme People’s Court unveiled the Faxin foundation model, the country’s first large legal language model, signaling rapid institutionalization of AI in the Chinese judiciary [para. 4]. Law, unlike medicine or finance, consists of highly structured, public texts—making it especially suited for training and deploying large language models (LLMs). Lawyers are heavy text consumers and researchers, so AI tools immediately enhance productivity [para. 5].

Numerous startups are capitalizing on this trend. Beijing Aiyu Intelligent claims to process over 10,000 lawsuits monthly relying on AI for every stage except court appearances, while in the United States, Harvey AI reached a $5 billion valuation just three years after launch. Industry giants like Thomson Reuters and LexisNexis are heavily investing in AI-related acquisitions and products [para. 6].

Despite the promise, significant risks persist. AI’s tendency to produce plausible errors—so-called “hallucinations”—has led to legal professionals submitting fabricated content to courts, sometimes resulting in sanctions or threats of criminal charges, as seen in the U.S., UK, and Australia [para. 7][para. 13][para. 14]. Nonetheless, AI’s role in early-stage research, statutory retrieval, and translation is growing, with some lawyers switching to faster, AI-powered tools for daily work [para. 10][para. 12]. For cross-border or complex cases, AI helps quickly map foreign legal frameworks, making legal inquiries more focused and efficient [para. 11].

Yet, the pitfalls necessitate caution: every AI-generated output is cross-checked against trusted databases like LexisNexis or Westlaw. Legal professionals stress that human skills—courtroom advocacy, negotiation, and error-spotting—remain irreplaceable, with AI mainly automating repetitive tasks [para. 15][para. 16][para. 17]. For some, automation is reshaping practice areas: for example, contract review demand is dwindling due to AI, while litigation work rises [para. 18].

A broad range of AI legal platforms, from U.S.-based startups to Chinese WeChat mini-programs, now generate tailored legal documents and research. Harvey, with 500+ enterprise clients and $100 million annual recurring revenue by August 2025, and LexisNexis, with global AI products, are prominent examples [para. 22][para. 23][para. 24][para. 25]. Costs for such tools can be substantial, but their time-saving benefits justify the expense for many firms [para. 27][para. 28]. Low-cost, all-purpose AI lawyer platforms remain aspirational; attempts at fully automated AI defenders in court have failed due to legal and regulatory barriers [para. 30].

China’s AI legal enterprises focus on modularizing tasks, with Aiyu Intelligent handling thousands of small-loan cases per month using minimal human oversight [para. 32]. PowerLaw AI targets contract review for small and midsize firms, shifting from lawyer-centric models to delivering immediate answers to businesses [para. 33][para. 34].

While the profession’s AI interest stretches back decades, digitization and searchable databases have been foundational to today’s advances [para. 37][para. 38][para. 39]. Chinese courts are leaders in judicial digitization, with hundreds of millions of rulings published online [para. 40]. Recent years saw the launch of digital assistants and the integration of LLMs into court workflows [para. 42][para. 43]. However, true “AI judges” remain distant: current consensus is that AI should only assist with repetitive or clerical work, leaving core judicial reasoning and discretion to humans, as confirmed by recent international surveys and leading professionals [para. 47][para. 48][para. 49][para. 50]. [para. 43][para. 47][para. 50]