Existing clients

Log in to your client extranet for free matter information, know-how and documents.

Client login

Staff

Mills & Reeve system for employees.

Staff Login
16 Nov 2026
6 minutes read

FutureProof: AI, litigation and the autonomous vehicle analogy

As businesses across the world continue to strengthen their policies around the use of artificial intelligence, the pressing question arises, who is ultimately responsible if something goes wrong? Is it the individual or the AI software itself?

The legal profession: The present position

The Solicitors Regulation Authority (SRA) “Misuse of AI Warning Notice”, published on 17 August 2026, clarified that AI has no separate legal personality and that solicitors using it remained accountable for their work and outputs. It also emphasised effective supervision, governance and human professional judgment.

The Bar Standards Board (BSB) “Guidance on the use of Artificial Intelligence and Other Technologies” explains how duties in the BSB Handbook apply to the use of AI. The onus remains on the barrister to be conscious about their use of AI and the way it may interact with the BSB ethical principles and Core Duties.

This overarching duty to act with independence means that AI can assist in making judgements and assessments, but it cannot become the main source of professional judgement. This duty requires lawyers to understand when technology may be influencing their professional decisions. Lawyers must make fully informed, critical and independent decisions.

A major flaw in AI systems is its ability to “hallucinate”. The Law Society in their practice note “Using artificial intelligence in legal practice” describes AI hallucinations as “incorrect or misleading results generated by AI systems”. Many lawyers have already been caught relying on fabricated legal authorities produced by AI which has led to individuals and firms facing both sanctions from the court and reputational damage.

A key example is the Ayinde case (R (Ayinde) v London Borough of Haringey; Al-Haroun v Qatar National Bank QPSC [2025] EWHC 1383 (Admin)). In this case, the claimant’s counsel relied on five cases that did not exist. The court suspected that AI was used to produce the false authorities, but it was disputed. As a result, a wasted cost order was issued and both solicitors and barristers involved were referred to their regulators.

Agentic AI and the autonomous vehicle

However, the question of responsibility becomes more complicated as AI advances beyond generating content based on individual prompts. The emergence of “agentic AI” represents a move towards systems carrying out a sequence of tasks in pursuit of an objective, with a greater degree of autonomy and less human intervention.

The Law Society’s article “What is agentic AI – and what does it mean for legal practice?” makes the distinction between generative AI (GenAI) and agentic AI. Whereas GenAI typically responds to individual prompts, agentic AI can work towards a broader goal by planning and carrying out a series of interconnected tasks with a greater degree of autonomy.

In a litigation context, the system could work through a series of steps, such as:

  • Reviewing and identifying documents for disclosure

  • Flagging potentially privileged material

  • Organising the relevant documents by issue

  • Preparing the first draft of the disclosure review document (DRD)

As useful as this may be, it poses the question as to whether it is sufficient for regulation to focus exclusively on the lawyer using the technology rather than the technology itself. As the distance between the human instructing and the AI performing increases, it creates greater pressure to consider the regulation of the AI system itself as well as those who develop and deploy it.

A good example of how this may work in practice is the automotive industry. Traditionally, the regulation of road use focused primarily on the individual behind the wheel. Then as vehicles became more automated, regulation shifted to address the automated driving system and placed responsibilities on the organisations creating and operating the systems.

Now, the newly enforced Automated Vehicles Act 2024 (AVA) creates a framework around self-driving vehicles. It sets a standard of what is classed as “self-driving” before any legal responsibility passes on from the human engaging with the vehicle. The explanatory notes for the AVA 2024 even go as far to say that “in certain circumstances where an authorised automated feature is engaged, an individual in the vehicle and in a position to operate the driving controls... may claim an immunity for offences arising from the way the vehicle is driven”. Is this a sign of things to come?

Limitations of the analogy

In litigation, it may not be possible for the lawyer to leave the driving seat to the extent that self-driving vehicles may do.

Litigation requires lawyers to exercise commercial judgement, consider ethical implications, exercise critical thinking and assess risk. Legal advice is inherently judgemental and contextual whereas a self-driving car can be measured against objective outcomes. It would be difficult to guarantee an AI system as “safe” in the same way.

A more likely development is a hybrid regulatory model. As AI continues to develop and assume more autonomous functions, regulation may begin to establish limitations and minimum expectations around the level of human oversight, testing, audit trails and the types of legal tasks that can be undertaken by AI. It will be necessary to continue to regulate professionals and firms, but also to create regulated AI governance processes and additional controls on high-risk legal AI applications.

Even if a hybrid regulation model is implemented, responsibility could lie at many points in the chain from the lawyer to the developer. If things go wrong, who would ultimately be responsible?

What can go wrong and who is responsible

AI could fail to identify a significant document, incorrectly categorise privileged material or produce an inaccurate summary of evidence. With GenAI, it may be easier to spot the errors earlier, but with the adoption of agentic AI it may be difficult to spot the error once the sequence of tasks has been carried out.

Additionally, using AI systems which are open source has significant implications for a lawyer’s duty to uphold the principle of client confidentiality.

In the immigration case, UK and R (Munir) v Secretary of State for the Home Department [2026] UKUT 81 (IAC), there were false AI generated authorities in the documents put before the court. The Upper Tribunal amended their judicial review claim form to include a requirement that legal representatives state that they have verified every cited legal authority. Furthermore, the judgment confirmed that referrals to the SRA and Information Commissioner's Office (ICO) will be made where confidential client information is put into a public AI system.

As these cases demonstrate, responsibility for a failure to properly verify or supervise AI-generated work currently lies with the lawyer and the firm, even if AI played a role in producing it. Things then become more complicated when a vulnerability in the systems cause confidential information to be exposed or if it behaves in an unexpected way. As AI becomes more autonomous, responsibility may therefore extend beyond where it is now, following a similar trajectory as the automotive industry.

Conclusion

AI is likely to play an increasingly significant role in legal practice, particularly as the technology develops from tools that assist with individual tasks towards systems capable of managing increasingly complex workflows. Agentic AI may represent the next stage of that development, with the potential to change not only how lawyers complete individual tasks, but how entire legal processes are managed.

The question is therefore becoming less about whether AI will form part of legal practice, and more about how it can be used effectively and responsibly. While the potential efficiencies are significant, appropriate human oversight, professional judgment and robust safeguards will remain essential.

For more information, please get in touch with Rajan Cheema or explore our FutureProof series.

Our content explained

Every piece of content we create is correct on the date it’s published but please don’t rely on it as legal advice. If you’d like to speak to us about your own legal requirements, please contact one of our expert lawyers.