Could lawyers be negligent for not using AI?

For the past few years, the warnings around lawyers using AI have been fairly predictable: do not rely on hallucinated cases, do not upload confidential client information into unsuitable systems, and do not assume that using AI shifts responsibility away from the lawyer.

However, there is a new question that is also worth considering.

Could a lawyer eventually be negligent for not using AI at all?

That possibility was raised recently by the UK Jurisdiction Taskforce (UKJT) latest statement, which suggested that if a professional acts negligently when using AI - or fails to use AI when it should have been used - they may face liability.

It sounds counterintuitive. Most of the legal conversation so far has focused on the dangers of using generative AI badly. But if AI becomes a normal and reliable part of professional practice, choosing not to use it may eventually attract scrutiny too.

The emerging position is that lawyers increasingly must make a competent and informed decision about whether AI should be used at all.

Ayinde and the responsibility to verify

That discussion comes against the backdrop of R (on the application of Ayinde) v London Borough of Haringey. 

The claimant brought judicial review proceedings after Haringey Council failed to provide interim accommodation while a homelessness decision was under review. The grounds of review, settled and signed by the claimant’s barrister, cited five entirely fictitious authorities, attributed passages to real cases that did not actually contain them, and materially misstated the effect of the relevant housing legislation.

Haringey’s solicitors uncovered the problem when they attempted to verify the authorities and requested copies, but none could be provided.

Whether the fictitious authorities had deliberately been included, or generative AI had been used to produce the material and an inaccurate account had then been given of its origins, the Court considered that the threshold for contempt was met either way.

The broader point is relatively simple: lawyers remain responsible, and indeed liable, for what they put before a court.

That responsibility is not displaced because material has been produced or assisted by generative AI. In much the same way that a supervising lawyer remains responsible for ensuring that junior work does not contain false cases or citations, lawyers cannot simply outsource verification to the technology.

Using AI does not outsource responsibility

The SRA has repeatedly warned that using AI does not “reduce, remove or replace” a solicitor’s regulatory or legal responsibility for the quality and accuracy of the service and information delivered to clients, courts, third parties or regulators.

Two risks are particularly obvious.

The first is relying on AI-generated authorities or legal propositions that are inaccurate or unverified. Ayinde demonstrates how serious the consequences can become when false material reaches a court.

The second is confidentiality. Inappropriately disclosing client information to an AI system may mean that the security of that information cannot reasonably be assured, creating obvious issues around client confidentiality and potentially privilege.

The UKJT identifies a broader range of possible failures too: using AI where it should not have been used, selecting an inappropriate system, or using AI without sufficient care.

In each case, the issue broadly is failing to ensure the contributions of AI are genuine aids rather than sources of misinformation.

There will therefore still be situations where the reasonable course is not to use AI.

Generally, where a tool is appropriately selected, its risks are understood and proportionate safeguards are in place, its use may be entirely consistent with the required standard of care. But where outputs are difficult to predict or errors difficult to explain, choosing not to use a particular tool may sometimes be the more professionally responsible decision.

The question remains one of reasonable professional judgment.

But what about not using AI?

In the UKJT’s July 2026 legal statement, they suggested that a professional could, in certain circumstances, be liable for failing to use AI.

Its reasoning is that AI is, ultimately, another professional tool. The relevant question should therefore not be whether AI was used, but whether a competent professional would reasonably have considered using it.

The UKJT gives the example of a solicitor acting in the Business and Property Courts on a case involving hundreds of thousands of documents.

If AI-assisted review is reliable, widely accepted and capable of making that exercise significantly more effective, could failing even to advise the client to consider using it eventually fall below the required standard of care?

The UKJT suggests that it could.

It compares the situation to a radiologist with access to an AI system that is extremely effective at identifying cancerous tumours. If the technology became a well-established professional tool, failing to use it could plausibly become harder to justify.

The same logic may eventually apply to legal work.

This does not mean lawyers are under a general duty to use whatever AI system happens to be available. Specific examples of this issue are yet to arise in the case law, and courts should be slow to conclude generally, with the benefit of hindsight, that a particular technology ought to have been adopted. 

Perhaps, then, AI is not reducing, removing or replacing the legal responsibility, but can be said to be altering its scope. The breach of duty stage of the negligence inquiry may be expanding to include the possibility of negligent failure to use these AI tools, where a competent professional peer with comparable rank would have done so.

The emerging development is the refining of a duty to make a competent and informed decision about whether AI should be used.

What should firms do now?

Moving forward, firms will increasingly adopt a cautious, risk-based approach to AI deployment. In terms of the accuracy and confidentiality concerns previously identified, in-house AI systems may certainly help mitigate the latter, but they of course do not entirely remove the need for proper due diligence. Professionals should assess whether an AI tool is suitable and appropriate for the task, including its reliability and security, particularly where the technology is new or innovative. As the UKJT notes, failing to undertake such checks may itself constitute a breach of professional duty.

Firms and chambers alike must carefully consider more than whether AI is being used, but whether the decision to use or not reflects the exercise of reasonable skill and care, and also ensure that the necessary verification and supervision is carried out overall to mitigate against liability.

Where this leaves lawyers 

AI does not seem to be changing the legal standard of professional negligence, but rather changing, within that standard, what competent professional practice will require.