Ross Baker, member of the Professional Indemnity Sector Focus Team (SFT) for FOIL and partner, Beale & Co, and Sam Zaozirny, member of FOL and senior associate, Beale & Co examines the complexity over the increased use of artificial intelligence (AI) and professional indemnity risks.

AI is now embedded in everyday professional work, from legal advice to clinical decision‑making. As it becomes part of routine practice, familiar questions of liability and responsibility are taking on new dimensions. For insurers and those handling professional risk, the key issue is no longer whether AI will influence claims, but how responsibility should be apportioned when an error arises from a process in which both human judgment and AI played a role.

AI as a tool not a tortfeasor

A persistent misconception in public and professional debate is that AI could itself be held liable for errors. In reality, AI cannot owe a duty of care, commit a tort, or make a misrepresentation. It is not a legal person. It is simply a tool, no different in principle from defective software, a malfunctioning medical device, or any other system used to support professional work.

Once this is understood, the analysis returns to first principles. If harm occurs, the key questions remain:

• Who selected the system?
• Who relied on it?
• Who was best placed to prevent the error?

AI is not legally exceptional. Courts are likely to approach it much as they would any other defective tool or technology, relying on established principles of contract, negligence and product liability.

Although the use of AI often sits in complex supply chains involving developers, vendors, integrators and users, disputes between those parties are generally determined by contract. The real difficulty emerges when harm is caused to someone outside that contractual matrix. There, the analysis returns to the core principles of negligence.

Professional Negligence: Same duties, new ways to breach them

AI does not lower the duty of care owed by professionals. Instead, it introduces new and increasingly common ways for that duty to be breached.

1. Using AI inappropriately – General‑purpose models are not designed for specialist or high‑stakes advice, as seen in the well‑publicised cases involving fictitious AI‑generated court citations.

2. Using the wrong tool for the job – Relying on consumer‑grade systems for regulated work, or models trained on flawed data, is likely to fall below the standard of a reasonably competent practitioner.

3. Lack of human oversight – Blind reliance on outputs, especially where they contradict evidence or common sense, are unlikely to be forgiven. In addition to their professional liabilities, directors and officers who use AI without exercising independent judgment may also find themselves breaching their own fiduciary duties to their company under s.173 of the Companies Act 2006. “The AI told me” will not operate as a defence.

4. Failing to use AI where expected – As certain forms of AI become clinical or professional standard, failing to use them may itself be considered negligent. The standard of care is likely to evolve alongside the technology.

The courts will not only ask whether a professional used AI, but how they used it; whether they understood its limitations, verified its reliability, supervised its outputs, and communicated their approach to clients. A failure to meet any of these obligations to the standard of a reasonable professional is likely to amount to negligence.

AI‑generated legal claims and the push for transparency

Where US courts have adopted mandatory disclosure rules, the English judiciary has taken a guidance‑led, rather than rule‑based, approach. The ‘AI Guidance for Judicial Office Holders’ issued in October 2025 emphasises the need for caution, transparency, and privacy when using generative AI tools in courts and tribunals. Court users are reminded that public chatbots can produce inaccurate or fabricated legal authorities, may reflect historic or foreign law, and must never be used for confidential material. The guidance also warns that litigants, especially those unrepresented, may rely heavily on AI tools, requiring courts to probe the source and reliability of submissions that appear AI‑generated.

Litigants in Person and the increasing use of AI

For litigants in person (LiPs), AI tools can level the playing field, helping them understand procedures and draft submissions; however, without the expertise to judge whether AI‑generated content is accurate, LiPs may inadvertently submit misleading or incoherent material. This can prolong proceedings, increase costs and damage their own cases.

AI‑generated claims introduce new complexities: pleadings may be longer, yet weaker, factual assertions may be unreliable, and it can be difficult to distinguish genuine argument from AI hallucinations. LiPs are also more emboldened in their arguments and less inclined to engage in settlement discussions.

Conclusion: AI may be new, but negligence isn’t

AI is not rewriting the law of negligence, but it is reshaping how those principles must be applied. Responsibility remains firmly with the human professional. Supervision, validation, documentation and transparent decision‑making will govern the threshold between competent use of emerging tools and a failure to meet the standard of care expected of a reasonable professional.

While the English courts have not followed the US in mandating disclosure of AI use, the judiciary’s recent guidance signals a clear direction of travel. As AI becomes embedded in professional practice, the legal and insurance markets will face new pressures: evolving standards of care, heightened expectations of oversight and an expanding cohort of claimants relying on AI without truly understanding it.

The technology may be new, but the central question for the courts remains unchanged: who was best placed to prevent the harm, and did they exercise reasonable care in doing so?