Hong Kong Law Firm AI Review: Limits and Liability

Capability limits, liability and the compliance reality

Attention to AI tools among Hong Kong legal practitioners rose markedly between 2024 and 2025. The Law Society of Hong Kong (HKLS) held a seminar on the impact of AI on the legal profession in August 2024 that drew more than 2,200 member registrations, and the record turnout reflects the profession’s collective attention to the topic. A 2025 LexisNexis survey showed that nearly 70% of Hong Kong legal practitioners were already using generative AI at work, with more than 60% using it weekly. But rising usage does not mean a clear compliance framework, and the gap between the two is the core challenge Hong Kong lawyers face in deciding whether to adopt AI.

HKLS’s Official Position

In January 2024 HKLS issued a position paper that positions AI explicitly as a tool, and not as a substitute for professional responsibility. The core principle it sets out: a lawyer bears full responsibility for the final quality of their work, and the adoption of AI tools must meet the requirements of ethical use and client protection. Technical competence is now regarded as a necessary skill for the modern lawyer, but the duty to adopt new technology does not rank above the duty to uphold professional standards.

In July 2024 the Hong Kong Judiciary issued its Guidelines on the Use of Generative Artificial Intelligence, which state explicitly that judicial functions cannot be delegated to AI, and that judges and judicial officers must keep independent judgement in any situation involving AI tools.

Where AI Is Actually Used in Legal Work

The most common uses of AI in Hong Kong legal practice concentrate on three types of task: contract review and summarisation, due diligence document analysis, and drafting of court documents. The best practice has two steps: AI produces a first draft, and a qualified lawyer verifies the sources, records the checking process and applies professional judgement.

Within this framework, AI’s role is an efficiency tool and not a replacement for judgement. AI can scan large volumes of documents in a short time and flag potential issues, but the final legal interpretation of contract terms, the risk assessment and the recommended strategy remain the lawyer’s responsibility.

The Legal Framework for Liability

Under Hong Kong’s legal framework an AI system has no legal personality and cannot be the bearer of legal liability. Under the principles of tort law, the users of AI tools, meaning law firms and individual lawyers, bear full legal liability for errors in AI-assisted work. A law firm must ensure that every piece of AI-generated work receives critical review by a qualified legal professional.

For accountability and verification design when AI output is wrong in production, see When an LLM Gets It Wrong: Accountability and Verification Design for Production AI.

Why AI-Assisted Work Must Be Recorded

Hong Kong courts treat the absence of written records as strong counter-evidence that a lawyer did not provide proper advice. If an AI-assisted workflow leaves no record of human review, it is hard to prove in a later dispute that diligent review actually took place.

Legal Professional Privilege (LPP) Risk

The HKLS position paper specifically warns that entering privileged communications into a third-party AI platform may amount to disclosure to a third party, carrying the risk of waiving legal professional privilege. For Hong Kong lawyers using cloud AI services this is a compliance issue that needs serious assessment: the data handling clauses in the terms of service should be reviewed carefully before any AI tool is adopted.

For the data flow and privacy questions to settle before handing documents to an AI system, see What Hong Kong Businesses Need to Know About Data Flows Before Using AI on Confidential Documents.

Documented Failure Cases

Two English cases give concrete warnings. In Ayinde v London Borough of Haringey [2024] EWCA Civ 1403, lawyers put five fabricated cases generated by AI before the court, including one claimed to come from the Court of Appeal, and the lawyers concerned were referred to their regulator. In Horoun v Qatar National Bank [2024] EWHC 912, a large number of citations were found to be fabricated or unrelated to the passages cited. In both cases the acting lawyers have been referred to regulators.

For a concrete citation verification workflow, see AI-Assisted Research: Hallucination Detection, Citation Verification and Source Tracking.

Summary

The reality of AI adoption in Hong Kong’s legal sector is that the tools are already widely in use while the regulatory framework is still taking shape. The HKLS position sets the minimum principle: AI is a tool, and responsibility rests with the lawyer. In practice three elements must be implemented together: complete records of human review, assessment of third-party AI platforms’ data handling terms, and independent verification of every AI-generated citation. Fabricated citations are the highest-risk failure mode documented so far in legal AI.

Further reading: AI Contract Translation: HK Legal Limits