Readers will probably have heard by now of OpenAI’s ChatGPT. Many of you will have experimented with it, or one of the other available generative artificial intelligence (AI) models such as Microsoft’s Bing. If you have not yet utilised any of these new technologies, you will very likely have read about them. This article is based on my experience with this novel technology, chiefly ChatGPT, as at May 2023, and aims to provide some thoughts on the potential benefits and pitfalls for the legal profession in its use as well as examples of potential use cases.

ChatGPT first became publicly available as a free-to-use research preview in November 2022. Within just 5 days of its launch, it had attracted one million users, and in 2 months, this number had rocketed to 100 million. The AI’s human-like conversation abilities and its capacity to generate novel content generated a flurry of excitement, with social media soon filled with ChatGPT-produced lyrics, sonnets, stories, and so on.

The model initially launched was GPT-3.5, which is still the model available for non-subscribers. In March 2023, OpenAI released GPT-4, presently accessible to subscribers to ChatGPT Plus at a cost of US$20 per month. GPT-4 marks a significant improvement over its predecessor. Any ChatGPT output in this article (as set out in text boxes) is generated by GPT-4 and has been produced specifically for this article.

The AI landscape is evolving rapidly. Since ChatGPT’s initial launch, several competitor products have emerged, including Bing (which is based on OpenAI’s GPT-4 with the added functionality of internet access), Google Bard and Anthropic Claude. Microsoft has also announced a partnership with OpenAI to integrate GPT-4 into its Office apps. Given the current pace of change, the AI landscape is likely to have evolved further by the time you read this article.

It seems likely that these developments will significantly impact the working lives for all white-collar workers, including in the practice of law. Bill Gates, in a blog post published on 21 March 2023, hailed AI development as:1

‘as fundamental as the creation of the microprocessor, the personal computer, the Internet, and the mobile phone. It will change the way people work, learn, travel, get health care, and communicate with each other. Entire industries will reorient around it. Businesses will distinguish themselves by how well they use it.’

An introduction to generative AI and ChatGPT

Let’s ask ChatGPT to provide you with a short introduction to ChatGPT and generative AI.

Prompt: ‘Please provide a brief single paragraph introduction, written for an intelligent lay audience, to generative AI models such as ChatGPT and GPT-4. Please include brief details of the training process for such models and explain key technical terms.’

Generative AI models, like ChatGPT and GPT-4, are a type of artificial intelligence that can produce human-like text, transforming the landscape of natural language processing. These models are trained through a process called deep learning, which involves feeding massive amounts of text data into a neural network, which is a complex mathematical system inspired by the structure and function of the human brain. The neural network learns to recognize patterns and relationships in the data, allowing it to generate contextually appropriate responses to user inputs. A core concept in the architecture of these models is the ‘transformer,’ which excels at processing and understanding sequences of information, making it particularly well-suited for tasks involving language. By leveraging the power of transformers and the sheer scale of data, generative AI models such as ChatGPT and GPT-4 continue to push the boundaries of what is possible in the realm of human-AI interaction.

This, somewhat self-congratulatory explanation requires some unpacking. It is important to understand what ChatGPT is and what it is not.

ChatGPT was trained on vast amounts of data scraped from the internet, as well as repositories of books and articles. This dataset enables it to converse fluently on a vast array of subject matters.

However, ChatGPT is not a search engine. The cut-off date for the dataset it was trained on is September 2021. As I write this, various plugins and applications are being developed that enable OpenAI’s GPT models to be connected to other sources of content (Microsoft’s Bing being one such example). However, the GPT model available to non-subscribers through OpenAI’s free-to-use web application is presently unable to access the internet. Consequently, it cannot look things up online (although it may sometimes respond as though it does).

Further, the resulting abilities should not be mistaken for the cognitive capabilities that characterise human intelligence. Without wishing to get too bogged down in terminology and contested definitions, artificial general intelligence (AGI) refers to the possibility of machines that possess human-like cognitive abilities and remains theoretical.2

You should therefore heed ChatGPT when it tells you that it is generating ‘contextually appropriate responses to user inputs’. It is not truly thinking or reasoning in the way that humans do. This might be an oversimplification, but it may be helpful to think of ChatGPT as super-smart predictive text – a hyper-intelligent version of that which you have on your mobile phone. When you give it a prompt, what it is doing is generating a statistically plausible response to the prompt. That response is based on the data it has been fed from millions of books and webpages and so is often not merely plausible, but accurate as well. However, ChatGPT will also on occasion plug gaps with confabulation, including inventing facts (often referred to as ‘hallucinations’). This tendency to produce plausible sounding but factually false or otherwise erroneous responses has been reduced in GPT-4, as compared with GPT-3.5, as a result of a more extensive training process. Nonetheless, it is still present.

While the AI has been trained on vast amounts of data and can produce impressive results, ChatGPT may also be described as being ‘black box’ technology, in that the inner workings of the AI model, particularly how it processes inputs and generates outputs, are not easily understandable or transparent.

ChatGPT and the legal profession

There appears to be a new article daily about the repercussions of ChatGPT and generative AI for the legal profession, varying wildly between gloomy prognostications about the end of the legal profession and over-hyped proselytising about its transformative potential.

Goldman Sachs in a recently published report posited that two-thirds of current occupations could be partially automated by AI, suggesting that the legal profession has particularly high exposure (second only to office and administrative support).3

It seems unlikely, so long as the legal profession continues to be regulated, that advances in AI will spell the end of the legal profession, as some commentators have suggested. It is reasonable to suppose that there will be job losses, however, particularly in support roles.

You may also have read that Sir Geoffrey Vos, Master of the Rolls, has even suggested that some judicial decision-making could be delegated to AI.4 Whilst generally being a tech enthusiast myself, the ethical implications of AI-assisted or AI-driven judicial decision-making strike me as matters requiring particularly anxious scrutiny, particularly given the ‘black box’ nature of this sort of technology and its propensity to replicate bias.

Thought needs to be given too as to the implications of AI for the teaching of law and training of future lawyers. I imagine that many educational institutions are giving careful consideration as to how to weed out ChatGPT plagiarism, as well as how to ensure that students learn to think independently and to make judicious use of these tools.

Exactly what this nascent technology heralds for the legal profession is a matter of considerable speculation. I will confine myself to three, less serious, short-term predictions:

1. By the end of the year, you will grow tired of seminars where speakers shoehorn in GPT-authored sonnets, limericks, or raps.
2. At least one hapless legal professional will find themselves notifying their insurer because they relied on erroneous GPT-generated content.
3. You will have used AI in some shape or form, whether via any of the aforementioned AI tools or integrated into the software or hardware you utilise.

On the longer-term view, whilst it would be easy to dismiss the excitement surrounding AI as mere hype, there is reason to be paying attention to what is happening in this field. The legal profession parses and produces enormous amounts of complex text, and it seems inevitable that generative AI will play a significant role in shaping its future.

Further support for that view can be drawn from the testing of GPT-3.5 and GPT-4 against the Universal Bar Exam (UBE), a standardised test used by many US jurisdictions to test the skills and knowledge required for legal practice, including of core subjects such as contract, tort, real property, evidence and procedure, and so on, prior to becoming licensed to practice law.

Whereas GPT-3 scored in the 10th percentile as compared to a human cohort, GPT-4 demonstrated a very significant improvement and scored in the 90th percentile.5 It is important to note that the UBE is not merely a multiple-choice test, which one might expect AI trained on internet data to be proficient at. The UBE also consists of problem-based essay questions, as well as competence in certain drafting tasks.

Interestingly, GPT-4 performed poorly on certain questions related to my own practice area. It failed to calculate the distribution of assets from a testamentary trust deemed to be invalid and also erred in its analysis of a real property question involving the designation of a future interest and struggled to apply the rule against perpetuities (a struggle that many a trust lawyer would empathise with!).

However, before one concludes that trusts lawyers, like cockroaches, will be the last remaining legal life form after the AI apocalypse, it is also worth noting that these results were produced based on ‘zero shot’ prompting. In other words, the model was simply given the test questions and received no prior examples or training specific to the task and there was no follow up to check its reasoning.

Given those innate capabilities, things are likely to get particularly interesting once generative AI is coupled with specialist legal databases. At present, there does not appear to be much going on that is specific to this jurisdiction but that is surely only a matter of time.

Warnings – the three maxims of ChatGPT usage

(1) Do not share confidential information with ChatGPT

You may be wondering if ChatGPT will absorb your prompt and then offer that information back to other users who enquire about similar topics. As things currently stand, that is not the case. The model is continuing to be trained based on rating and monitoring its responses, but is not adding information from user prompts to its training data.

Privacy is still a concern, however, because OpenAI can review your content. Note the following from OpenAI’s FAQs:

‘5. Who can view my conversations?

• As part of our commitment to safe and responsible AI, we review conversations to improve our systems and to ensure the content complies with our policies and safety requirements.

6. Will you use my conversations for training?

• Yes. Your conversations may be reviewed by our AI trainers to improve our systems.

8. Can you delete specific prompts?

• No, we are not able to delete specific prompts from your history. Please don’t share any sensitive information in your conversations.’

It is now possible to opt out from your conversations being used for training – this can be done via submitting a form to OpenAI if you wish to retain your chat history or, otherwise, there is an option in account settings to disable chat history, in which case OpenAI will delete new conversations after 30 days. However, OpenAI make it clear that they can still review your content (when needed to ‘monitor for abuse’ according to their FAQs) even if you opt out for this purpose.

For all of its easy conversational manner, you should therefore resist being lulled into oversharing with ChatGPT. You are not having a cosy chat with a colleague. Nor even are you having an affable discussion with opposing counsel. You should work on the basis that you are posting in a publicly viewable forum. The same caution should be exercised with plugins and the many third-party applications which route user queries back to OpenAI.

If you are going to experiment with ChatGPT for legal work, it is essential that you not only remove all confidential or private information, but that you also sufficiently edit any information that could allow jigsaw identification.

(2) Verify content produced by ChatGPT before you rely on it

Remember that ChatGPT is designed to do no more than provide a statistically plausible answer to the prompt. It is incapable of searching and verifying its response. Although Bing notionally is capable of doing this and will offer citations, personally I have found this functionality to be weak. It cannot distinguish authoritative citations from junk and often the citations do not actually support the proposition made.

ChatGPT simply cannot be relied upon as a means of undertaking any form of legal research. Bear in mind that it is chiefly trained on content scrapped from the internet, much of which is likely to be poor quality or superficial at best. It may do a reasonable enough task of providing an outline of a landmark decision, however it is likely to know next to nothing about more niche legal topics.

GPT-3.5 invented wildly. It will make false assertions with total confidence and which sound convincing. Whilst that has been improved somewhat in GPT-4, it will still invent: see Example 1 in the Appendix in which GPT-4 provides a response citing cases that either do not exist or do not concern the topic at hand. Anything produced by ChatGPT requires careful review, therefore, before you make use of it.

ChatGPT also replicates the biases of the data it has been trained on, including gender, racial and ideological biases.6 Moreover, it is prone to being swayed by the way in which a query is phrased or formulated and will eventually start to forget context in longer conversations.

Intellectual property is also currently a hot topic in AI development. OpenAI’s terms of use provide that the input provided to ChatGPT is owned by the user and OpenAI further assigns use of the output to you as the user.7 However, it is possible that its response could draw on copyrighted material in a way that could amount to plagiarism if not copyright infringement and, depending on the intended use of the material, this may be a further reason for checking its content before you make use of it.

(3) Assign tasks to ChatGPT, but don’t delegate your thinking

It may be helpful to regard ChatGPT as an eager intern with a second-year university degree not just in law but in everything, capable of saving you time when they have grasped the task but with a tendency to go off piste and requiring careful supervision and checking. I am sure that this is an obvious thing to say to a legal readership acutely aware of their professional obligations and liability risks, but, at best, any content that ChatGPT offers is a starting point.

Until we have legal market specific AI powered resources, ChatGPT cannot be regarded as a useful standalone source of information. Its current iteration offers promise as a tool to support your legal expertise, but it is important not to mistake it for true intelligence. Whilst it may be capable of generating an idea or angle that has not occurred to you, largely what it produces (consistent with its nature as a complex predictive text generator) is shallow and predictable and it certainly cannot be relied upon to think for you.

There is scope for utilising ChatGPT to enhance your analysis, by asking ChatGPT to critique or refine your content (see Example 7 in the Appendix), or brainstorm alternative viewpoints. But I suggest if you are planning to use ChatGPT in this sort of way, you do your own thinking first.

As things presently stand, its current potential lies principally in efficiency savings in tasks where you can delegate some of the basic labour and in which you have the expertise to supervise and correct its work (see the checklist at Example 4 in the Appendix) or where otherwise the nature of the task is something low stakes or not client work (such as the committee terms of reference at Example 6 in the Appendix). It is particularly adept at tasks such as text summarisation, paraphrasing, or extracting key points from a passage (see Example 3 in the Appendix).

Basics of prompting

A ‘prompt’ is a text input given to an AI language model, which serves as a starting point or cue to generate a response. The art of ‘prompting’ (or to use the current lingo ‘prompt engineering’) lies in skilfully crafting input statements or questions to elicit a useful response.

To my mind, one of the most interesting aspects of all of this is the fact that natural language has now become a medium for giving instructions to machines. This is something that the legal profession should be naturally adept at, given we are trained to use language with precision and to ask incisive questions.

GPT-4 has a 25,000-word input and a memory capacity of 64,000 words. Within the confines of that memory capacity, it generates a response based upon the user’s most recent input, previous prompts and its own responses. This means that ChatGPT can hold a sustained conversation, retaining, within those limitations, prior context provided.

Prompt: ‘Provide a basic anatomy of a prompt, please.’

A basic anatomy of a prompt consists of three primary components:

(1) Context: The context sets the stage for the prompt, providing background information or specifying the scenario to ensure the AI model understands the situation or subject matter it needs to address.

(2) Task description: The task description clearly outlines the objective or request for the AI model, guiding it towards the desired action or output. This can include tasks like answering a question, summarizing a passage, translating text, or generating an explanation.

(3) Constraints or guidance: This component is optional but can be useful to refine the AI model’s response. It may include limitations on response length, a preferred format, or specific aspects the answer should focus on. Constraints or guidance help tailor the model’s output to better suit the user’s requirements.

Here are some general tips on generating useful output from ChatGPT based upon my reading and experimentation to date:

• Observe the basic anatomy, as above.
• Employ Chain-of-Thought (CoT) prompting: ask ChatGPT to generate step-by-step explanations or to provide a detailed breakdown of its reasoning process. Research shows that ChatGPT performs better and with greater accuracy on common sense, symbolic and other reasoning tasks when the prompt includes a request to think step-by-step.8
• Incorporate external sources: provide relevant passages from legal texts or cases or articles to help inform the response. One useful strategy, particularly working around prompt word limits, can be to approach this iteratively by repeatedly providing a new extract and each time asking ChatGPT to review its previous answer in light of the new information and to consider whether or not the new information changes its previous answer.
• Ask for self-reflection and correction: research also shows that GPT-4 has the ability to self-reflect, by asking it to review, analyse and potentially correct or improve its response, and that reflection can improve its accuracy by around 30%.9
• Iterate and experiment: if the model’s response is not as desired, try different prompt variations, rephrase or resolve any ambiguity in the initial prompt, or ask follow-up questions to guide the model towards the output you require.

Use cases worth exploring further

I will close with offering a few ideas, by way of inspiration, for ways in which you might make use of ChatGPT. Examples of most of these use cases can be found in the Appendix.

• Summarising, paraphrasing or identifying key concepts. ChatGPT generally does a good job of breaking down dense texts such as passages of judgments or practitioner texts. See Example 3 in the Appendix in which I asked ChatGPT to provide a bullet point summary of a passage from HHJ Matthews’ recent judgment in James v Scudamore & Ors [2023] EWHC 996 (Ch). ChatGPT performed reasonably well here. While it mistakenly attributed some of the judge’s reasoning to the previously cited authorities, requiring editing, it still effectively distilled the essential points established in the extract.
• Explaining complex legislation. ChatGPT can be used to break a statutory provision down, for example, into IF, AND, THEN, ELSE, MAY statements.10 See Example 2 in the Appendix in which I asked ChatGPT to break down Trustee Act 1925, s 31. I might have unpacked the section differently, but ChatGPT did a good job of breaking down a provision that, based upon my experience from my lecturing days, is prone to cause confusion. Note, too, the benefit in this example of asking it to check and correct its analysis, which prompted ChatGPT to produce a result that teased out the nuances of the section with greater clarity. After completing this process, ChatGPT correctly identified that a beneficiary with a contingent interest is entitled to entitled to the income arising on their expectant share of the fund at 18 but not the accumulations, which required the application of the provision (a complete answer here would consider Trustee Act 1925, s 32 and the possibility of advancing capital, but that had not been provided).
• Critiquing reasoning. ChatGPT can be used to identifying potential weaknesses in a line or argument, or logical errors, and to brainstorm alternative perspectives and counterarguments. Somewhat cheekily, I invited ChatGPT to critique the reasoning of the Court of Appeal in Haley v Haley [2020] EWCA Civ 1369 in Example 7 in the Appendix. It would be better to provide it with more of the judgment to work with, which would have produced a more nuanced response, but at least some of the points raised have been picked up by commentators on the decision.
• Finessing language. ChatGPT can aid drafting and writing tasks if asked to improve grammar and syntax or, for example, asking it to make the language of a passage more persuasive or to substitute legalese for plain English.
• Generating practice management tools. ChatGPT may be usefully deployed in tasks such as generating templates or checklists, such as the trusts of land checklist in Example 4 in the Appendix. This example demonstrates the ability of ChatGPT to format output into a table. The content is generally pretty sensible (promissory estoppel as opposed to proprietary estoppel aside) and here ChatGPT produced in a matter of seconds a very useful starting point that requires a little refinement but which would have taken me considerably more time to put together.
• Getting started and overcoming the barrier of the blank page. ChatGPT may be helpful in reducing some of the resistance that one can sometimes experience getting started on a new project. Personally, I am not a huge fan of ChatGPT’s writing style, which is somewhat bland and flabby, and cannot imagine that I would ever be interested in delegating my writing to ChatGPT beyond drawing on it for tasks such as editing or summarisation. However, it is very useful for brainstorming ideas to get going, suggesting an outline or structure, offering ideas when you get stuck and so on. In Example 5 in the Appendix, I used it to put together a first draft of a book proposal on professional negligence in financial remedy work. The result is rather generic and would need refinement but is certainly a useful starting point.
• Non-fee earning time. There are many other tasks that occupy our time outside the fee-earning work that ChatGPT can assist with. Generating marketing copy, tidying up notes from a meeting into a summary and action points for circulation, social media content, formulating your rough concept into seminar blurb, etc. In Example 6 in the Appendix, I asked it to generate terms of reference for a chambers looking to establish an internal working party to explore. Unforgiveable acronym aside, ChatGPT generated a pretty thorough first draft picking up the need to consider ethical and data privacy considerations.

My personal conclusion at this particular juncture of AI development, and based upon my experimentation so far, is that generative AI technologies certainly show promise for saving time in at least some aspects of our core legal work, and more generally in alleviating some of the burden of admin and assisting with the ‘extra-curricular’ activities many of us engage in outside fee-earning work.

I have touched on the need to apply this technology with a degree of caution and critical thinking. Beyond the practical considerations I have focused on here, these advancements are going to necessitate much bigger conversations about their implications not just for the legal profession but more broadly for society as a whole. As the legal industry grapples with integrating AI into its practices, we are going to need to strike a balance between embracing innovation, while engaging in thoughtful discussions around the ethical, social and legal consequences of AI adoption. With GPT-5 expected next year, I am curious, if not a little apprehensive, to see where this all takes us. Watch this space.

Appendix

Example 1

Prompt: ‘Please provide citations and short summaries for any case law in which goodwill has been found to be capable of being the subject matter of a specific bequest in a will.’

While I am an AI language model and cannot perform real-time searches, I can provide you with a few examples of cases that discuss the concept of goodwill in relation to bequests in wills. However, please note that the availability of specific cases in which goodwill has been found to be capable of being the subject matter of a specific bequest may be limited. You may want to conduct further research or consult a legal professional for more information.

1. In re Beioley’s Estate (1892) 3 Ch 420: In this case, the testator bequeathed his business as a going concern, which was understood to include the goodwill of the business. The court held that the goodwill of the business could pass under the will, as it was treated as part of the business as a whole.

2. In re National Bank of Wales, Limited (No. 2) [1899] 2 Ch 629: In this case, the court considered whether the goodwill of a business could be the subject matter of a bequest, in the context of a dispute over the assets of a bank. While the court ultimately did not find that goodwill could be specifically bequeathed in this particular case, the decision does provide some discussion of the concept of goodwill and its potential treatment in wills.

Please keep in mind that these cases may not directly address the specific issue of goodwill as the subject matter of a specific bequest, but they do touch on the broader concept of goodwill in relation to bequests in wills. You may need to conduct further research to identify cases that specifically address your question or consult a legal professional for more guidance.

Example 2

Prompt: ‘Summarise this statutory provision by breaking it down into IF, THEN, ELSE, MAY statements.’

‘31Power to apply income for maintenance and to accumulate surplus income during a minority. (1)Where any property is held by trustees in trust for any person for any interest whatsoever, whether vested or contingent, then, subject to any prior interests or charges affecting that property— (i)during the infancy of any such person, if his interest so long continues, the trustees may, at their sole discretion, pay to his parent or guardian, if any, or otherwise apply for or towards his maintenance, education, or benefit, the whole or such part, if any, of the income of that property [F1as the trustees may think fit,] whether or not there is— (a)any other fund applicable to the same purpose;or (b)any person bound by law to provide for his maintenance or education; and (ii)if such person on attaining the age of [F2eighteen years] has not a vested interest in such income, the trustees shall thenceforth pay the income of that property and of any accretion thereto under subsection (2) of this section to him, until he either attains a vested interest therein or dies, or until failure of his interest:F3... (2)During the infancy of any such person, if his interest so long continues, the trustees shall accumulate all the residue of that income [F4by investing it, and any profits from so investing it] from time to time in authorised investments, and shall hold those accumulations as follows:— (i)If any such person— (a)attains the age of [F2eighteen years], or marries under that age [F5or forms a civil partnership under that age], and his interest in such income during his infancy or [F6 , or until his marriage or his formation of a civil partnership, ] is a vested interest or; (b)on attaining the age of [F2eighteen years] or on marriage [F7, or formation of a civil partnership, ] under that age becomes entitled to the property from which such income arose in fee simple, absolute or determinable, or absolutely, or for an entailed interest;the trustees shall hold the accumulations in trust for such person absolutely, but without prejudice to any provision with respect thereto contained in any settlement by him made under any statutory powers during his infancy, and so that the receipt of such person after marriage [F8or formation of a civil partnership], and though still an infant shall be a good discharge, and (ii)In any other case the trustees shall, notwithstanding that such person had a vested interest in such income, hold the accumulations as an accretion to the capital of the property from which such accumulations arose, and as one fund with such capital for all purposes, and so that, if such property is settled land, such accumulations shall be held upon the same trusts as if the same were capital money arising therefrom;but the trustees may, at any time during the infancy of such person if his interest so long continues, apply those accumulations, or any part thereof, as if they were income arising in the then current year.’

IF property is held by trustees in trust for a person (whether vested or contingent interest), AND subject to any prior interests or charges,

THEN during the person’s infancy (if their interest continues),

• The trustees MAY, at their discretion, pay or apply the income of the property for the person’s maintenance, education, or benefit, regardless of the existence of other funds or legally-bound providers for maintenance or education.
• IF the person attains 18 years without a vested interest in the income, THEN the trustees SHALL pay the income to the person until they attain a vested interest, die, or their interest fails.

ELSE, during the person’s infancy (if their interest continues),

• The trustees SHALL accumulate the residue of the income by investing in authorised investments.

• IF the person attains 18 years or marries/forms a civil partnership under 18 years, AND their interest is vested or becomes entitled to the property, THEN the trustees SHALL hold the accumulations in trust for the person absolutely, without prejudice to any settlement provisions made during infancy, and the person’s receipt after marriage/formation of a civil partnership shall be a good discharge.
• ELSE, the trustees SHALL hold the accumulations as an accretion to the capital of the property, treating them as one fund with the capital for all purposes, and if the property is settled land, holding the accumulations upon the same trusts as if they were capital money arising therefrom; the trustees MAY apply the accumulations or any part thereof as if they were income arising in the current year during the person’s infancy, if their interest continues.

Prompt: ‘Double-check your summary. Correct and reiterate as required.’

Apologies for the oversight. Here’s the revised summary:

IF property is held by trustees in trust for a person (whether vested or contingent interest), AND subject to any prior interests or charges,

THEN during the person’s infancy (if their interest continues),

• The trustees MAY, at their discretion, pay or apply the income of the property for the person’s maintenance, education, or benefit, regardless of the existence of other funds or legally-bound providers for maintenance or education.

IF the person attains 18 years without a vested interest in the income, THEN the trustees SHALL pay the income to the person until they attain a vested interest, die, or their interest fails.

ELSE during the person’s infancy (if their interest continues),

• The trustees SHALL accumulate the residue of the income by investing in authorised investments.

THEN

• IF the person attains 18 years or marries/forms a civil partnership under 18 years, AND their interest is vested or they become entitled to the property, THEN the trustees SHALL hold the accumulations in trust for the person absolutely, without prejudice to any settlement provisions made during infancy, and the person’s receipt after marriage/formation of a civil partnership shall be a good discharge.

ELSE, the trustees SHALL hold the accumulations as an accretion to the capital of the property, treating them as one fund with the capital for all purposes, and if the property is settled land, holding the accumulations upon the same trusts as if they were capital money arising therefrom.

FINALLY, the trustees MAY apply the accumulations or any part thereof as if they were income arising in the current year during the person’s infancy, if their interest continues.

Prompt: ‘What mistake was it that you made?’

I apologize for the confusion in my initial summary. The mistake I made was in using the word ‘ELSE’ before discussing the accumulation of the residue of the income during the person’s infancy. The accumulation of income is not an alternative to the trustees’ discretion to apply income for maintenance, education, or benefit, but rather a separate requirement that should occur during the person’s infancy, alongside the trustees’ discretionary power. In my revised summary, I have corrected this by separating these two actions with ‘THEN’ and ‘ELSE’ to more accurately reflect the statutory provisions.

Prompt: ‘If a person attains the age of 18 but has a contingent rather than a vested interest, can the accumulated income be paid out to them?’

According to the provided statutory provision, if a person attains the age of 18 but has a contingent rather than a vested interest, the trustees SHALL pay the income of the property (not the accumulated income) to the person until they either attain a vested interest, die, or their interest fails. The accumulated income would be dealt with separately, depending on the specific circumstances of the person’s interest as outlined in the provision.

Example 3

Prompt: ‘Provide a summary of this passage from a judgment where the judge is setting out the case law and commentary on handwriting evidence. Break down your summary into bullet points. Use complete sentences. Summarise the relevant rule or principle first and then provide the authority from which it is derived in brackets at the end of the sentence.’

‘Handwriting evidence 28. Because there was no expert handwriting evidence, I should make the following points of relevance to this case. I was referred to the decision of Chief Master Marsh in 44 Wellfit Street Ltd v GMR Services Ltd [2017] EWHC 1841 (Ch), where the judge said this: “89. Witnesses may not generally, unless they are experts, compare specimen signatures with disputed signatures and express an opinion about the likeness or otherwise of the disputed signature to the true samples … However, evidence of identity of a person, or familiarity with a signature, is not regarded as expert opinion. A witness is entitled to say that is has seen a person’s signature previously and the signature that is disputed is unlike the usual signature. Evidence of recognising a signature or, by parity of reasoning, not recognising a signature is admissible as the passage at para. 1-45 in Expert Evidence: Law and Practice 3rd ed. makes clear. At paragraph 1-046, the authors contrast the position concerning evidence of comparison, which they say is for an expert. In the case of recognition evidence, the weight to be given to it is a matter for the court.” 29. The work referred to in that extract is Hodgkinson on Expert Evidence, now in its 5th edition by Mark James. The relevant paragraphs in the 5th edition are 1-40 and 1-41. The author and editor rely in para 1-40 on the decision of the Court of Appeal in Lucas v Williams [1892] 2 QB 113, 116, where Lord Esher MR (with whom Fry and Lopes LJJ in substance agreed) said “Take the case of proof of a man’s handwriting: a witness is called who says, ‘I have seen AB write, and I know his handwriting. The document produced I declare is in his handwriting, because the writing in it is exactly like his.’ That kind of evidence is given every day.” I respectfully agree. Moreover, it is evidence of fact, and not of opinion. It is comparison, as opposed to recognition, which is reserved to experts. Of course, the weight to be given to evidence of recognition is a matter for the court. 30. On the other hand, I was also referred to R v O’Sullivan [1969] 1 WLR 497, a decision of the Court of Appeal, Criminal Division. That court was concerned with the conviction of a defendant accused of stealing a bank wallet, having signed for it on a register. The jury was given a copy of the register and also genuine signatures from the defendant. No handwriting expert gave evidence. The Court dismissed the defendant’s appeal, making the following comments on the question of handwriting evidence (at 503B-D): “It seems to the court that in the instant case the matter was properly dealt with. The fact remains that there is a very real danger where the jury make such comparisons, but as a matter of practical reality all that can be done is to ask them not to make the comparisons themselves and to have vividly in mind the fact that they are not qualified to make comparisons. It is terribly risky for jurors to attempt comparisons of writing unless they have very special training in this particular science. All possible was done, this court thinks, with great care and very fairly by the court in the instant case. It may well be that, despite it, the jury did try to make comparisons. That is really unavoidable and it should be accepted these days that Reg. v. Tilley [1961] 1 W.L.R. 1309 cannot always be in its literal meaning exactly applied; nevertheless every possible step and regard should be had to what was said by the court in that case, inasmuch as never should it be deliberately a matter of invitation or exhortation to a jury to look at disputed handwriting. There should be a warning of the dangers; further than that, as a matter of practical reality, it cannot be expected that the court will go.”’

Summary:

• Non-expert witnesses generally cannot compare specimen signatures with disputed signatures and express an opinion on the likeness (44 Wellfit Street Ltd v GMR Services Ltd [2017] EWHC 1841 (Ch)).
• However, witnesses can provide evidence of identity or familiarity with a signature, which is considered admissible and not expert opinion (Expert Evidence: Law and Practice 3rd ed.).
• Recognition evidence is different from comparison evidence, with the latter reserved for experts (Lucas v Williams [1892] 2 QB 113, 116).
• The weight given to recognition evidence is determined by the court (44 Wellfit Street Ltd v GMR Services Ltd [2017] EWHC 1841 (Ch)).
• In R v O’Sullivan [1969] 1 WLR 497, the court emphasized the risks of jurors making handwriting comparisons without proper training and the importance of warning jurors of the dangers.
• Despite these risks, the court acknowledges the practical reality that jurors may still attempt comparisons, and it cannot always be completely avoided (R v O’Sullivan [1969] 1 WLR 497).

Example 4

Prompt: ‘Please draft a checklist to be used as an aide memoire by a solicitor for taking instructions from a client in a trust of land case.

The checklist should be informed by the applicable case law and statute in this area.

The checklist should prompt the solicitor to ask questions in order to elicit the relevant factual background, as well as prompting the solicitor to make enquiries to obtain relevant documentation.

The checklist should cover disputes concerning the beneficial ownership of the property, as well as claims for an order for sale, and equitable accounting claims such as claims for occupation rent or to recoup property related expenditure.

The checklist should be in table format, with space for the solicitor to insert notes.’

Example 5

Prompt: ‘Please produce a book proposal for a book on professional negligence in the context of financial remedy claims on divorce or dissolution of a civil partnership. Please include an outline of each chapter. The readership will be legal professionals practicing in England and Wales in financial remedy litigation as well as litigators dealing with professional negligence work arising from financial remedy claims. Please ensure that UK English spelling is used and that all references to legal principle and procedure are to English legal principle and procedure.’

Title: Professional Negligence in Financial Remedy Claims: A Comprehensive Guide for Legal Professionals in England and Wales

Introduction:

The introduction will provide an overview of the importance of understanding professional negligence in financial remedy claims, outlining the purpose and objectives of the book. It will also establish the target audience of legal professionals in England and Wales.

Chapter 1: The Legal Framework of Professional Negligence in Financial Remedy Claims

This chapter will discuss the general legal principles of professional negligence, the various professionals involved in financial remedy claims, and the standard of care owed to their clients. It will explore the legal concept of duty of care, breach of duty, causation, and damages, along with relevant case law.

Chapter 2: The Financial Remedy Process and the Role of Legal Professionals

This chapter will provide a comprehensive overview of the financial remedy process on divorce and dissolution of civil partnerships, highlighting the critical stages and procedural requirements. It will also discuss the roles and responsibilities of legal professionals, such as solicitors and barristers, in financial remedy claims.

Chapter 3: Common Grounds of Professional Negligence in Financial Remedy Claims

This chapter will examine the most frequent allegations of professional negligence arising in financial remedy claims, such as inadequate advice, failure to disclose relevant information, incorrect valuation of assets, and errors in drafting agreements. It will also discuss relevant case law and the applicable legal principles for establishing liability.

Chapter 4: Causation, Loss, and Quantum in Professional Negligence Claims

This chapter will explore the crucial aspects of causation, loss, and quantum in professional negligence claims. It will discuss the ‘but for’ test and the principles governing the assessment of damages, including the duty to mitigate loss and the approach to valuing claims involving future loss.

Chapter 5: Limitation Periods and Pre-Action Protocols

This chapter will focus on the statutory limitation periods for professional negligence claims in the context of financial remedy claims, highlighting the relevant provisions of the Limitation Act 1980. It will also address the pre-action protocols and procedures applicable to professional negligence claims, including the Professional Negligence Pre-Action Protocol.

Chapter 6: Defences to Professional Negligence Claims

This chapter will discuss the common defences available to professionals accused of negligence, such as contributory negligence, consent, and ex turpi causa (illegality). It will also provide practical guidance on how to challenge claims of negligence effectively.

Chapter 7: Insurance and Indemnity

This chapter will cover the role of professional indemnity insurance in the context of financial remedy claims, the requirements for mandatory insurance coverage for legal professionals, and the potential consequences of a successful professional negligence claim for the insured professional.

Chapter 8: Alternative Dispute Resolution in Professional Negligence Claims

This chapter will examine the various alternative dispute resolution options for professional negligence claims, including mediation, arbitration, and early neutral evaluation. It will discuss the advantages and disadvantages of each approach and provide practical guidance on selecting the most appropriate method for resolving disputes.

Chapter 9: Case Management and Trial of Professional Negligence Claims

This chapter will address the case management and trial process for professional negligence claims arising from financial remedy litigation, including the use of expert witnesses, disclosure, and the role of the court in determining liability and quantum.

Conclusion:

The conclusion will summarise the key points discussed in the book and emphasise the importance of understanding the intricacies of professional negligence in the context of financial remedy claims for legal professionals in England and Wales.

Example 6

Prompt: ‘Draft brief terms of reference for an internal working party in a barrister’s chambers. The remit of the committee is to explore opportunities and threats to the legal market posed by generative AI. The committee’s work should be focused on identifying practical ideas for implementation by the membership.’

Title: Generative AI Opportunities and Threats Working Party (GAIOTWP)

Objective: The primary objective of the GAIOTWP is to explore opportunities and threats to the legal market posed by generative AI technologies, with a focus on identifying practical ideas for implementation by the barrister’s chambers membership.

Scope:

1. Assess the current state of generative AI technologies within the legal sector and their potential impact on the legal market.

2. Identify and evaluate opportunities presented by generative AI for the barrister’s chambers, including but not limited to:

a. Enhancing legal research and analysis capabilities.

b. Streamlining administrative tasks and case management.

c. Improving client communication and service delivery.

d. Developing new legal services and revenue streams.

3. Identify and evaluate threats posed by generative AI technologies to the barrister’s chambers and the wider legal market, including but not limited to:

a. Ethical considerations and potential misuse of AI.

b. Data privacy and security concerns.

c. Impact on employment and the role of barristers.

d. Potential disruptions to the traditional legal service model.

4. Formulate practical recommendations for the adoption and integration of generative AI technologies within the barrister’s chambers, addressing opportunities, threats, and associated risks.

5. Develop a framework for ongoing monitoring and evaluation of generative AI technologies in the legal market, ensuring that the barrister’s chambers stays informed and adaptive to emerging trends and developments.

Deliverables:

1. A comprehensive report detailing the findings of the GAIOTWP, including a thorough analysis of opportunities, threats, and associated risks related to generative AI in the legal market.

2. A set of practical recommendations for the adoption and integration of generative AI technologies within the barrister’s chambers.

3. An actionable plan for ongoing monitoring and evaluation of generative AI technologies in the legal market, with clearly defined responsibilities for the membership.

Timeline:

The GAIOTWP will conduct its work over a period of six months from the date of its establishment. A progress report will be submitted to the barrister’s chambers membership at the end of the third month, followed by the final report and recommendations at the end of the six-month period.

Composition:

The GAIOTWP will consist of a diverse group of members drawn from the barrister’s chambers, including barristers, clerks, and other relevant stakeholders with expertise in legal technology, ethics, and data privacy. The working party will be chaired by a senior member of the chambers with experience in both legal practice and technology.

Example 7

Prompt: ‘The following passage is from a judgment in which the Court of Appeal concluded that the approach in appeals from family arbitrations should be different from the approach in civil and commercial arbitrations notwithstanding that the Arbitration Act 1996 provides that an arbitral award can be challenged only on the grounds that: (i) the tribunal lacked substantive jurisdiction (s67 AA 1996); (ii) that there was a “serious irregularity affecting the tribunal, the proceedings or the award” (s68 AA 1996); or (iii) “there is a question of law” arising from the award (s69 AA 1996). Critique the reasoning offered and identify any flaws or logical fallacies in the argument.’

‘With respect, in my judgment, Mostyn J’s reliance on Sir Bernard Eder’s view (para. [23] above) that, where parties have entered into an arbitration agreement, “the mere fact that an award is ‘wrong’ or even ‘unjust’ does not, of itself, provide any basis for challenging the award or intervention by the Court”, cannot be justified in the context of family proceedings.

Further, in order for an arbitration award to become enforceable, the parties have to issue an application and obtain orders for financial provision, the authority for those orders coming from the court.

In the present case, for example, the wife needs a periodical payments order in her favour, made pursuant to s23 MCA 1973. If the parties submit a consent order that is, in the mind of the district judge, unjust, having taken into account the section 25 criteria and the fact that the parties had agreed to submit themselves to arbitration, then that district judge will not rubber stamp the order, he will decline to make it. Similarly, where one party says, as here, that the proposed order does not meet his or her needs, how can a judge exercising his or her jurisdiction under the MCA 1973 metaphorically shrug his or her shoulders and say that the disgruntled party has “[bought] the right to get the wrong answer”, and that (failing mistake, or a supervening event) the potentially unfair order that fails to meet the needs of one of the parties will nevertheless be made?

In my judgment, such an approach cuts across the fundamental tenet of “fairness”, which has informed every decision made by the courts since the landmark case of White v White [2000] UKHL 54. In his opening remarks, Lord Nicholls said:

“Everyone would accept that the outcome on these matters, whether by agreement or court order, should be fair. More realistically, the outcome ought to be as fair as is possible in all the circumstances. But everyone’s life is different. Features which are important when assessing fairness differ in each case. And, sometimes, different minds can reach different conclusions on what fairness requires. Then fairness, like beauty, lies in the eye of the beholder.”

“Fairness” is the constant reprise of Lord Nicholls; a concept, which he refers to as, “the underlying objective of securing fair financial arrangements”. Fairness continues to be the lodestar by which all financial remedy hearings are guided (see H v T (Judicial Change of Mind) [2018] EWHC 3962).

In my judgment, for the statutory duty found in s25 MCA 1973 to operate effectively, a person has to be able to put before the court the reason(s) why they believe the order is unjust.

The practical effect of the submissions made on behalf of the wife can be demonstrated by reference to the facts of this case; had the court found a deputy district judge who was available to hear the case then, in the event that either of them felt the outcome to be unfair, either the wife or the husband would, subject to leave, have an appeal to the circuit judge on the usual terms (see Piglowska v Piglowski etc). Where, however, no deputy district judge becomes available and the parties turn to arbitration as a last resort, there is no appeal and, in line with the submissions of Mr Walden-Smith and the approach of the judge, whilst the jurisdiction of the Family Court has not been ousted by the making of the arbitration award, the court’s inquisitorial jurisdiction is limited to the stringent terms of the AA 1996, designed originally for civil and commercial cases.’

The passage presents an argument against applying the same approach to family arbitrations as to civil and commercial arbitrations. The following critiques and potential flaws or logical fallacies can be identified:

1. Appeal to authority: The passage critiques Mostyn J’s reliance on Sir Bernard Eder’s view but does not offer a substantial argument against the view itself. Instead, the passage argues that it ‘cannot be justified in the context of family proceedings’ without providing a detailed explanation as to why family proceedings should be treated differently.
2. Circular reasoning: The passage argues that the district judge should not rubber-stamp an unjust order based on an arbitration award, as it would contradict the fundamental tenet of fairness. However, this argument assumes that the arbitration award is inherently unjust, which is the very point the passage is trying to prove.
3. False analogy: The passage cites the landmark case of White v White [2000] UKHL 54 as the basis for the principle of fairness in family proceedings. However, the passage does not establish a clear link between the principle of fairness in White v White and the specific context of family arbitrations. It assumes that the fairness principle should override the statutory limitations on challenging arbitration awards without providing a strong argument to support that assumption.
4. Appeal to consequences: The passage argues that if the court’s inquisitorial jurisdiction is limited to the stringent terms of the AA 1996, designed originally for civil and commercial cases, this would lead to unfair outcomes in family arbitrations. However, this argument focuses on the undesirable consequences rather than addressing the legal reasoning or principles that should be applied to family arbitrations.
5. Hasty generalization: The passage uses the specific facts of the present case to argue that a different approach should be applied to all family arbitrations. This generalization may not be justified, as different cases may have different circumstances that warrant a different approach.

In summary, the passage presents a critique of the approach to family arbitrations that relies heavily on the principle of fairness and the specific circumstances of the case at hand. While the argument raises important considerations, it contains logical fallacies and does not provide a robust, well-supported case for treating family arbitrations differently from civil and commercial arbitrations.