Artificial Intelligence (AI) in Litigation: Navigating Professional Liability for Malaysian Practitioners

Across dispute resolution practice, AI tools are being used to synthesise legal research, analyse evidence, produce transcriptions, and generate first drafts of cause papers and submissions. The efficiency gains are real, but so are the legal consequences that follow. Malaysia has not been a passive bystander. The Sabah Magistrates’ Court made regional history on 19 February 2020 when it employed the Artificial Intelligence in Court Sentencing (“AiCOS”) System, the first AI-assisted sentencing tool in Asia. At the governance level, the Government launched the National AI Office (“NAIO”) in December 2024, and the Malaysian Bar established its Legal Tech, AI and Sandbox Committee to guide practitioners. This article examines how AI is reshaping courts and legal practice in Malaysia, what professional obligations it generates, and how the Malaysian position compares with Singapore, the United Kingdom and other countries.

AI in the Courts: Sentencing and Judicial Discretion

Malaysia’s most significant judicial AI deployment is the Sentencing Calculator pilot in Sabah and Sarawak, applied to wildlife offences under the Wildlife Conservation Act 2010. By processing large datasets of prior sentencing outcomes, the system proposes consistent penalties and seeks to harmonise penalties and reduce sentencing disparity across regional courts. In PP v Supani Jumal [2025] MLRHU 3441, the High Court held that AI sentencing calculators are non-binding aids and do not constitute ‘written law’ within the meaning of Article 160 of the Federal Constitution. This position was solidified in PP v Kamil Suteh [2025] MLRHU 2570, where the court reaffirmed that trial judges exercise unfettered judicial discretion; as algorithmic suggestions carry no binding weight and may be departed from without explanation. Ultimately, judicial AI in Malaysia was confirmed by the courts to be only an advisory instrument: one that may inform, but cannot replace, the exercise of judicial discretion.

The Malaysian Bar’s Position

The Malaysian Bar addressed AI in legal practice through two circulars that together constitute the operative regulatory framework.

Bar Council Circular No. 342/2023 (24 November 2023)

It warns against core risks of using generative AI (ChatGPT) such as hallucinated cases, fabricated citations, confidentiality breaches and algorithmic bias. Its central and enduring instruction was for lawyers to independently verify all AI output against traditional legal databases before any reliance.

Bar Council Circular No 242/2025 (3 July 2025)

It supplements but does not replace the 2023 circular. It broadens the scope from ChatGPT specifically to all generative AI tools, introducing recommended and prohibited use lists, and introduces the 4R decision framework alongside a risks-and-mitigation table.

Ultimately, both circulars establish an unwavering standard: AI must never be relied upon blindly, independent verification is compulsory and that the lawyer bears the ultimate professional responsibility for the work.

AI in Legal Practice: Drafting, Supervision and Client Data

AI drafting tools are increasingly used for standard-form pleadings, cause papers and correspondence. The critical point under Malaysian professional conduct rules is not the use of AI itself, but the lawyer’s supervision of its output. As AI becomes pervasive, the standard of the ‘reasonably competent practitioner’ will be reinterpreted accordingly: a competent practitioner today is expected to understand the limitations of the AI tools they deploy, including the risk of hallucinations.

A further obligation arises when client material is fed into a third-party AI platform. Where a platform’s standard terms of service permit training on user inputs, uploading client documents without consent risks an unauthorised disclosure. In R (Munir) v Secretary of State for the Home Department [2026] UKUT 81 (IAC), the UK Upper Tribunal held that uploading client documents to an open-source AI platform places them in the public domain, breaching confidentiality and waiving privilege, and that supervising partners bear full responsibility for a junior’s use of AI tools with client data. The convenience of the tool does not diminish the accountability of the lawyer; express client consent and appropriate data processing agreements are required before any such upload takes place.

The Danger of AI Hallucinations: Global Precedents and Sanctions

The professional liability risk from AI hallucinations is acute and well-documented. AI fabricates legal content in ways: inventing cases with plausible names and real-looking citations that simply do not exist; fabricating quotes attributed to real cases; and misrepresenting real authorities, twisting them to support a point the case does not actually stand for. According to the AI Hallucination Cases Database (D. Charlotin, as of 14 September 2026), there are now 3,124 documented court decisions involving AI hallucinations globally, of which 814 involve practising lawyers and 2,542 concern fake or misrepresented case law. The trend is not slowing.

The pattern recurs across jurisdictions. In Mata v Avianca, Inc, 678 F Supp 3d 443 (SDNY 2023), attorneys were fined USD 5,000 each and ordered to notify every falsely-named judge after submitting six (6) ChatGPT-fabricated citations. In Zhang v Chen (BC Supreme Court, 2024), the lawyer was ordered to personally pay the opposing side’s costs of unravelling two (2) ChatGPT-invented authorities. Closest to home, in Tajudin v Suriaya (Singapore High Court, 2025), counsel’s written submissions cited a case that did not exist; when opposing counsel flagged it, counsel first called it a “clerical error” and blamed a junior, before admitting it was AI-generated and fictitious. Personal costs of S$800 were awarded against the counsel. The court rejected the attempt to blame the junior and was of the view that the senior had a duty to supervise the work of his junior colleague.

In the most recent illustration of the problem, a federal judge in Mississippi removed and fined all four lawyers on both sides of a civil trial; barred two of them from that court for two years. This incident was after counsel on each side filed AI-generated citations to non-existent cases. One lawyer admitted she was “unaware that AI could produce hallucinated cases and explained that she did not even know what a hallucinated case was”. The proceedings were cancelled entirely. As the judge found, a partner who disregards her own firm’s AI verification policy engages in an “extreme dereliction of professional responsibility” and ignorance of AI’s limitations is no defence.

The English Administrative Court’s warning in R (Ayinde) v London Borough of Haringey; Al-Haroun v Qatar National Bank [2025] EWHC 1383 (Admin) is to the same effect: counsel’s duty extends to verifying every cited authority regardless of source. In Malaysia, any unverified AI-generated citation placed before a court would expose counsel to disciplinary proceedings under the Legal Profession (Practice and Etiquette) Rules 1978 and to adverse costs orders. Circular 342/2023 set the rule from the outset, that practitioners owe a duty to independently verify all output against the CLJ, MLJ or equivalent before it is filed or relied upon.

Comparative Governance: Singapore, The United Kingdom and The United States

Singapore

Singapore offers the most instructive regional model. The Singapore Court issued Registrar’s Circular No 1 of 2024 (23 September 2024) – Guide on the Use of Generative Artificial Intelligence Tools by Court Users and the Ministry of Law published its Guide for Using Generative AI in the Legal Sector in March 2026, both expressly anchoring AI use to practitioner duties under the profession acts. In private practice, major law firms have adopted Harvey AI for legal research, drafting, and document review; and the Small Claims Tribunals (SCT) has incorporated Harvey AI to summarise case documents for Tribunal Magistrates and individuals representing themselves in the SCT. Tajudin v Suriaya shows that Singapore courts are equally prepared to impose personal costs consequences on counsel who fail the verification standard, a direction Malaysian courts may follow as AI-related filings increase.

 

United Kingdom

In the United Kingdom, Cork v Smith [2026] EWHC 1199 (Ch) and R (Ayinde) (supra) together establish that supervisors who wave through AI-generated content without verification incur personal liability, and courts are prepared to examine the precise prompts typed and the steps at which verification failed. As of today, the UK does not have any AI-specific regulation or legislation covering AI as a technology. The SRA Warning Notice (2023) and Bar Standards Board (BSB) guidance operate through the existing professional conduct framework, holding practitioners personally accountable regardless of the tools used.

 

United States

In the United States, accountability flows primarily through Rule 11 of the Federal Rules of Civil Procedure — as applied in Mata v Avianca and the Mississippi case, which makes AI literacy a component of the competence duty under the rules of professional conduct. Individual federal courts have also issued standing orders requiring disclosure of AI use in filed documents.

Conclusion

The global litigation landscape up to mid-2026 demonstrates that courts no longer view AI hallucinations or data leaks as mere technical accidents. Instead, the failure to supervise, audit and verify machine-generated output is universally treated as a direct breach of a lawyer’s professional responsibility. Rather than waiting for a precedent, given that Malaysia has yet to see its first formal penalty for AI hallucinations, the immediate burden falls on law firms to establish internal auditing frameworks and strict tech-compliance protocols. Ultimately, while AI tools offer undeniable efficiencies in processing data, they cannot replace human intellect, ethical judgment or an advocate’s paramount duty to the court. To safeguard both client privilege and professional standing, Malaysian law firms must enforce strict internal verification protocols and remember that the signing practitioner always owns the file.

This article was authored by Suaran Singh Sidhu (Partner) and Cheryl Chung (Associate).

---

Key contacts

Please contact our team if you have any questions.

Suaran Singh Sidhu | Head, Dispute Resolution & TMT