Taskforce warning
Lawyers who fail to use artificial intelligence (AI) could find themselves on the wrong end of a negligence claim, according to a new legal statement from the UK Jurisdiction Taskforce (UKJT).
Published this week, the taskforce’s Legal Statement on Liability for AI Harms notes that although there is no AI-specific liability regime in England and Wales, existing English law is capable of determining whether and when a person may be liable for harm caused by the use of AI. Among its conclusions is that lawyers can be liable not only for using AI incorrectly, but also for failing to use it at all.
Whether a solicitor or barrister has breached their duty by failing to use AI, according to the statement, depends on whether “a reasonable professional of a comparable rank/specialism” would have used it in the same situation, a test that will draw on regulatory guidance as AI adoption becomes more established across the profession.
Among the examples given is a solicitor in the Business and Property Courts who fails to advise their client that an AI-assisted tool could be used to review a large volume of documents.
Other professions are also addressed. The statement cites a radiologist who does not use a cost-effective, accurate AI tool for identifying tumours, and an auditor who fails to deploy AI to detect anomalies across transaction volumes too large for manual review, as further examples of potential liability for not using AI.
Comparable space is given to liability for using AI incorrectly, with the taskforce listing several scenarios: failing to carry out due diligence on an untested tool, failing to explain to a client how the AI works, entering confidential material into an insecure system, and failing to identify errors or bias in the output.
The statement also addresses AI chatbots that communicate on a company’s behalf, concluding that liability is likely to arise where a business holds a chatbot out as speaking for it. There is no English case law on the point yet, though the statement notes that a Canadian court has held an airline liable for false statements made by its chatbot.
Operating under LawtechUK and chaired by Sir Geoffrey Vos, the Master of the Rolls, the UKJT is a government-backed body tasked with clarifying how English law applies to new technology. Its previous statements on cryptoassets and smart contracts have been adopted directly by judges.
Commenting on the statement, Matthew Lavy KC, chair of the UKJT’s drafting committee, said: “English law has long been capable of rising to the challenges of novel technologies, and we consider that it will be able to do so in relation to AI.”
The statement was published as several firms continue to invest in their own AI tools. Shoosmiths has rolled out Project Apollo, a contract review tool developed with Microsoft, Freshfields has partnered with Anthropic on legal AI tools, and Kirkland & Ellis has set aside $500 million to build its own platform.
Who determines when an AI is “accurate”, particularly given its propensity to hallucinate?
It seems unfair that a lawyer who “fails to advise their client that an AI-assisted tool could be used to review a large volume of documents” will be deemed to have been in breach of a supposed “duty” to use AI, when at the same time, and as has been seen to occur in real life, if the same review tool misses some key piece of information within the documents, the lawyer will also be in breach if they fail to spot such an error? Then the lawyer may as well have done it themselves, with the confidence of their own skills over AI’s!
The lawyer is then stuck in a situation where they are at the mercy of an AI being accurate, as if they don’t use them, they’re in breach, and if they do use them and it gets things wrong, they too are in breach…
The technological echo chamber that is AI lacks the human quality of having common sense. I pity our system if we establish ourselves so that lawyers become mere “error identifiers” of AI’s output, rather than having both the ability and liberty to think for ourselves!
It’s a tool that when used in the right way, can lead to tasks being completed faster, but yes, lawyers as always need to ensure it is accurate.
There’s nothing inherently wrong with this.
Are lawyers still telling clients to physically drag all their files into a warehouse for the other side to carry out due diligence (as they did back in the 90s), or are they making use of tech to run a digital dataroom, being efficient and saving costs for the client – as this is the current standard of legal service?
Same goes for AI. If you have an AI system that can pre-review a million documents in 3 days, are you telling your client that using the AI system to do a pre-review will prevent your firm having to deploy 40 paralegals on it for 2 months?
And of course you have to ensure it is accurate. You are the lawyer using a tool are you not? It is still your process, your advice, your work, your client. Why shouldn’t you be held accountable?
I did not argue that lawyers shouldn’t be held accountable for inaccuracies due to their use of AI. Lawyers must be accurate in their own work, AI or otherwise. Of course, if AI is used in the correct way, it can be helpful and speed up processes.
Instead, the issue is that lawyers are deemed to have to use AI purely for the sake of supposed “efficiency”, when such AI systems may not be efficient at all if they spurt out errors which go undetected by lawyers. This has been demonstrated to have happened in many cases, such as with Sullivan & Cromwell, among others. In such cases, I would much rather a firm I was using have had taken the “2 months with 40 paralegals” (let’s keep the employment for our paralegals, too! They should be the foundation of our long term lawyer-development, rather than ropey AI tools!) and have produced a robust, and most importantly, correct court submission, than to have instead very quickly produced hallucinated rubbish!
I am sick to death of AI being presented as an absolutely necessary tool – of course it can be helpful when used correctly, but I should hope lawyers can keep some confidence in their own abilities, and be able to use them without having to be shackled to an AI output!
Frankly, I do not think this is accurate. And looking into this taskforce, it is clear that they are trying to promote their own goals, rather than give an unbiased interpretation of the situation.
I do not think the law has developed to the point that AI has now become viewed as an essential tool, so much so that the lack of use of it would be considered negligence. Frankly preposterous!
What a contradiction: we are constantly warned- about the fears of future lawyers becoming too dependent on Ai !!
If you use something like casewax then it just help you
The interesting part for me is that the real question isn’t simply whether lawyers should use AI, but how they use it.
For confidential case files, I’d be much more comfortable with an AI tool that runs locally and shows exactly which document and page an answer came from. That makes it much easier for the lawyer to verify the output rather than treating the AI’s answer as the conclusion.
That’s actually why we built CaseWax a local AI tool for Windows that searches a firm’s own documents without uploading them to a cloud AI service, with the source file and page attached to each answer.
AI shouldn’t replace the lawyer’s judgement. It should make the tedious parts easier to check.