By Ross Risby, David Kwok and Julie Wong
|
Published 28 September 2026
Following on from our recent article on the insurability of AI risks [link to article], in which we explored whether existing policy wordings (in particular cyber policies) are equipped to respond to AI-related losses, this article considers the likely impact of AI on professional indemnity insurance.
Professionals are increasingly incorporating AI tools into their practice with some predictions that lawyers, accountants, consultants and other professionals may be replaced entirely by AI in the future. Although suggestions of the demise of these professions may be overstated, AI is already changing how professionals operate, from document review and research to report preparation, information management and decision-making support.
More autonomous forms of AI misuse may present a significant future risk, particularly in light of the recent OpenAI Hugging Face breach, which showed how quickly AI-enabled systems can identify and exploit vulnerabilities at scale. However, the more immediate concern for professionals, and their insurers, appears to be the reliance on unverified AI-generated material that transpires to be inaccurate, incomplete or entirely fabricated.
Recent cases, such as R (Ayinde) v Haringey London Borough Council [2025] EWHC 1383 (Admin), involving lawyers relying on fabricated AI-generated authorities demonstrate that AI hallucinations remain a real risk. Whilst such tools may improve efficiency, and are proving highly attractive to firms coming under increasing pressure to deliver ahead of time and at reduced cost, they also create new pathways for professional error.
Jurisdictions across Asia are taking steps to address the risk of AI hallucinations in court proceedings, largely by applying existing professional duties to the use of AI rather than creating a separate liability regime. In practical terms, lawyers, litigants, and judges remain responsible for ensuring that AI-assisted material is accurate and supported before it is put before the court.
Singapore has already taken enforcement action through costs sanctions where fictitious AI-generated authorities were cited, while Hong Kong courts have made clear that both represented parties and litigants in person remain responsible for the accuracy of AI-assisted submissions.
The overall regional approach is not to prohibit the use of AI, but to require proper verification, disclosure where necessary and clear accountability so that hallucinated authorities, unsupported factual assertions and unreliable AI-generated evidence do not undermine legal proceedings.
However, despite the regulatory safeguards being put in place there remains a very real risk that a professional or firm could be exposed to a claim arising out of an AI hallucination. The question then arises: would such a claim be covered under a traditional professional indemnity policy?
Professional indemnity policies currently tend not to explicitly cover AI-related errors or omissions, although these claims are likely to be analysed as conventional negligence claims. Provided the claim falls within the definition of professional services and no exclusion applies, many PI policies may respond irrespective of whether AI formed part of the factual background.
The more difficult question may be whether existing policy language is sufficiently clear where the relevant service has been delivered, checked or influenced by AI, particularly in circumstances where professional services are increasingly delivered through models involving limited professional involvement. Reports of an SRA-regulated law firm in the UK facilitating claims without solicitor involvement illustrate why insurers are likely to focus closely, before inception, on how AI-enabled services are supervised and disclosed.
Policyholders should assess how AI is being used in their business, what external AI exposures they face, whether their current insurance programme responds to those exposures, and whether their proposal form disclosures remain accurate. Similarly, insurers are increasingly reviewing their policy wordings, risk assessment processes and pricing and focussing their disclosure requirements on how AI-enabled services are supervised and disclosed to ensure that they can properly understand and assess the risk presented.
AI is already changing the delivery of professional services. The key risk for firms is not simply that AI may produce inaccurate output, but that such output may be relied upon without adequate human review or clear responsibility for verification. Against that background, policy wordings are likely to come under closer scrutiny. Insurers can be expected to ask more detailed questions about the use of AI in proposal forms, while policyholders will want greater clarity on whether losses involving AI fall within, or outside, the cover they have purchased. Those insurers able to take a clear and commercially realistic position on AI-related cover are likely to be better placed as demand develops.
