Ad/Marketing Communication
This legal article/report forms part of my ongoing legal commentary on the use of artificial intelligence within the justice system. It supports my work in teaching, lecturing, and writing about AI and the law and is published to promote my practice. Not legal advice. Not Direct/Public Access. All instructions via clerks at Doughty Street Chambers. This legal article concerns AI Law.
In late 2024, the Master of the Rolls, Sir Geoffrey Vos, gave two very intriguing speeches both focussing on AI in the law. Sir Geoffrey Vos seems to adopt a similar view to Lord Justice Birss, which I discussed in this blog post previously.
That view is that lawyers and judges must ready themselves for the sweeping changes AI will bring. One speech was given in February 2025 at the LawtechUK Generative AI Event, and the other on 9 October 2024. They can be read here and here.
I will just briefly summarise each speech.
Speech at the LawtechUK Generative AI Event (5 February 2025)
Embrace AI or Risk Obsolescence? Sir Geoffrey Vos emphasised that the legal profession has no meaningful option to avoid AI. Other industries have embraced it at all levels, and law must follow suit.
He focuses on the notion that lawyers will handle disputes over both the misuse and the missed-use of AI. This includes improper use of AI tools and potential negligence in not utilising AI where it could have improved service delivery.
Another prominent theme is how AI can expedite and cheapen legal advice, making justice more accessible. Sir Geoffrey advocates a Digital Justice System in which routine or lower-value disputes are resolved online, potentially aided by AI.
He points to four key pillars in the UK Judiciary’s AI guidance:
- Know what generative AI does or doesn’t do.
- Never insert confidential information into public LLMs.
- Always remain accountable. Any AI output you use is still your responsibility.
- Fundamental Human Rights
Sir Geoffrey Vos underscores that human rights laws and regulatory tools may need rethinking in the face of advanced AI. An urgent debate is required about which decisions should remain a human prerogative and which we might entrust to AI.
The focus is on the legal world moving forward with AI prudently and ethically, but with confidence that it will unlock vast benefits.
Speech at the Irish Law Society Event (9 October 2024)
Sir Geoffrey delved into how the GDPR, especially Article 22, and the EU’s AI Act can restrict or forbid fully automated decision-making that significantly impacts individuals’ rights unless there is consent or legislative authority.
He highlights potential claims around using public data to train large language models. Getty Images vs Stability AI is a leading example of the copyright debate. The re-use of “copyleft” material also raises tricky questions about licensing obligations for AI-generated code. That was one of the three cases considered in this article.
He warns against over-regulating AI before fully understanding its capabilities. Premature legal barriers might hamper beneficial adoption, yet fundamental rights must be preserved.
In essence, the Master of the Rolls here again offers a balanced position, urging caution in the face of regulatory complexities and emphasising the importance of not stifling innovation.
Comment
Initially, what caught my attention was the consistent message from Sir Geoffrey Vos and Lord Justice Birss regarding AI’s inevitability and transformative potential (provided we maintain ethical safeguards and accountability). There appears to be no significant difference between them on the fundamentals. Each acknowledges that AI will, in time, handle tasks traditionally reserved for humans. Yet there are some nuances in emphasis between their speeches worth noting. Particularly in terms of timeline, Lord Justice Birss considers actual AI-based decision-making may become commonplace “in the not-too-distant future” for certain types of claims. Nevertheless, both judges are ultimately aligned in their belief that AI will continue to enhance the justice system, thereby demanding robust vigilance regarding rights and responsibilities.
However, for the purposes of this post, I was particularly struck by Sir Geoffrey Vos’s comment that one of the biggest fields of legal activity in the coming years will likely be claims arising from “the negligent or inappropriate failure to use AI.” This raises intriguing questions about the circumstances under which an organisation might become liable if it deliberately avoids adopting AI. One example that immediately comes to mind is the medical profession. Could doctors soon face claims if they opt not to use AI-driven diagnostic tools capable of detecting illnesses earlier than traditional methods? Similarly, might businesses face legal action for failing to utilise AI to promptly identify financial fraud or cybersecurity breaches?
I then started to think closer to home. Is this something that lawyers themselves should start to consider more seriously? Might solicitors and barristers soon face claims for not using AI tools capable of rapidly analysing vast volumes of legal material, thereby potentially avoiding costly oversights or missed authorities? If a solicitor relies solely on traditional research methods, could that be considered negligent in the future, especially when sophisticated AI systems can now uncover relevant cases or statutory interpretations in mere moments? Even the judiciary itself might face scrutiny if their decisions fail to reflect insights or patterns readily apparent to AI. It’s provocative to imagine a scenario where the legal profession, traditionally cautious and methodical, might soon need to justify not adopting AI as part of everyday practice. The concept seems almost counterintuitive, yet perhaps inevitable, prompting deeper reflection on the evolving meaning of professional competence in the age of AI.
But I’m not sure this is entirely right. I’ve been researching and using AI tools extensively, both professionally and personally, for some time. No matter how competent I’ve become at prompting and assimilating data from AI, I don’t always save much time, in fact, it can often take longer. As Sir Geoffrey Vos rightly observes, confidential data should never be entered into a large language model (LLM), which means there must always be a careful sifting process to ensure confidential information isn’t accidentally exposed or generated. Moreover, as I hope has been clear from this blog, AI hallucinations are a genuine concern. Every output must be carefully scrutinised, as inaccurate or completely fabricated information can send even the most conscientious lawyer down a rabbit hole, diligently cross-checking legal databases to verify whether a seemingly perfect case actually exists. Brief summaries generated by AI can also miss pivotal facts because in my experience AI cannot yet reliably determine what is relevant and what isn’t. The overall result can sometimes be more time spent rather than saved.
Additionally, an often overlooked issue is the cybersecurity risk associated with adopting AI. Increased reliance on AI systems can inadvertently expose sensitive client information or create new vulnerabilities, especially if these tools are not securely integrated into existing IT infrastructure. This further supports the need for cautious, thoughtful integration of AI rather than rapid, blanket adoption.
Yet, despite these reservations, perhaps there’s another dimension worth exploring which is our client’s expectations. As clients become increasingly aware of AI’s capabilities, could they soon expect lawyers to integrate these technologies regardless of their current limitations or inefficiencies? I have often wondered whether clients will start selecting legal services based on the AI they use. If so, this might accelerate the adoption of AI, placing further pressure on the profession to keep pace. Furthermore, it’s important to acknowledge that while today’s inefficiencies may seem significant, AI technology is evolving rapidly. Regular reassessment of these arguments will become essential as these tools mature and improve in reliability.
Additionally, the regulatory landscape will likely adapt to reflect these emerging responsibilities. Current discussions implicitly assume the existing legal frameworks will hold, but it’s plausible, perhaps unavoidable, that professional guidelines or regulatory bodies will soon introduce clearer standards specifically governing AI’s use in legal practice. Such standards could shape how negligence in the context of AI adoption is defined.
For these reasons, I believe claims related to failing to use AI may not be straightforward. Courts are likely to develop nuanced tests determining whether adopting AI is genuinely beneficial and reasonable under particular circumstances. Ultimately, the best outcome may come from lawyers working closely with AI rather than seeing it as something to fully adopt or avoid. Everyone has their own way of working, and similarly, different AI tools suit different people. Just like lawyers, no two AI models are exactly the same. The future of law might therefore depend on each professional choosing the right AI partner, creating a varied landscape shaped by personal preferences and practical needs.
