Beyond “Be Careful” Josh McBride on Tautoru AI 

Heard about Tautoru AI, but not quite sure what it is or why it matters? We spoke with barrister Josh McBride about the website he built, the problem it is trying to solve, and why lawyers need clearer guidance on using AI safely.

For barristers who haven’t come across Tautoru AI before, what problem are you trying to solve?

Tautoru AI was created to address a practical gap for lawyers: helping them make informed decisions about whether and how to use AI tools, and what guidelines they need where sensitive material is involved.

The problem is the complexity of working out what you can and can’t put into an AI chatbot. People are told to be careful, which is right, but that blanket caution doesn’t really help unless they know what ‘careful’ actually looks like. Tautoru is about giving people practical guidance on that - helping them understand what may be safe to input, what may not be, and how to take a sensible and defensible position on AI use.

There are plenty of AI tools on the market. Why are you focused only on general-purpose AI tools and not on the legal-focused ones?

The distinction between “public” AI tools and “legal” AI tools is not as clear as many people assume. Most legal AI products are still powered by the same frontier LLM models – such as those from OpenAI, Anthropic, Google, and xAI – usually through an API and with a legal-specific “wrapper” sitting over the top. It’s the same technology.

Lawyers therefore need to understand the basics of the underlying LLMs and how these are offered to the public through chatbot “wrappers” like Claude or Chat GPT, not just the branded legal tool in front of them. 

A legal AI product may look different from, say, Google’s Gemini chatbot product, and be more contained, but that does not necessarily make it safer or better. 

The real questions are what LLM powers the tool, where the data goes when you interact with it, what the subscription or enterprise terms allow, and whether material could be retained, reviewed, or reused. These are difficult questions requiring careful consideration.

The issue is not only whether data is used for training the underlying LLM. Some providers also reserve the ability to review prompts and outputs for safety or policy reasons. For lawyers, that matters because sensitive file material, particularly in areas such as criminal work, may be escalated for human review. Whether that might breach professional or ethical duties is something lawyers need to actively consider. 

As a rule, the frontier models such as ChatGPT or Claude are generally much safer than lawyers realise, provided the right sort of subscription is selected and lawyers actively consider what sorts of material they intend to input into it.

Can lawyers realistically assess AI products themselves, or is the market now too technical?

Most lawyers understand that AI matters, but many find it overwhelming.

What I’m seeing on the ground is that 5% of lawyers are really embracing it, but 95% aren’t. The 5% that do, are starting to accelerate away.

Lawyers are unsure what is safe and are already under enough professional pressure. This market is technical, and becoming more so, but that doesn’t mean lawyers can step away from it. 

We’re seeing that at the moment, where lots of lawyers are complaining about AI-generated material from clients, and wanting it to stop. But if lawyers won’t use AI, the clients will. Nature hates a vacuum. So I tell my clients, please don’t provide me with your AI generated materials. I prefer to use the tools myself, and I will do it in a way that hopefully will deliver a better output than what they’ve come up with.

The point is not that every lawyer must become an AI enthusiast, but that ignoring the tools will not stop clients using them. In fact, the opposite.

What is the long-term aim of Tautoru AI?

Tautoru AI isn’t a commercial product. It is a side project. It’s a place to put my thinking on AI, and a resource people can look at when they want to understand the tools in more practical terms.

It’s free. It’s not anything that anyone has to pay for.

The long-term aim is modest. I want the site to remain useful and up to date while the issues stay live, but he does not suggest it will become a paid platform. In fact it may have a limited shelf life. It is a resource for the current stage of the conversation.

What concerns you most about how the profession is reacting to AI?

The risk is that the profession responds with blanket warnings and prohibitions, rather than practical guidance. If lawyers are simply told not to use public AI tools, that will not stop the technology being used. 

It will just mean people use it quietly, without clear standards around safety, privilege, subscription models or human review.

The problem with prohibition is that all it will do is effectively drive use underground. Of course, lawyers and clients are still going to use it. They’re just not going to tell you that they are. This creates all sorts of problems.

This is the biggest change for our profession in its history. And it’s accelerating. Lawyers are cautiously starting to think about how to use AI and how to deploy it safely and properly. I don’t know that there’s really much choice. If we are to remain relevant, the profession needs to understand how to use AI tools, and how to use them safely.

Where should lawyers start?

For lawyers unsure where to begin, start small. Choose one major platform and use it first for mundane, low-risk tasks involving non-sensitive material: summarising a judgment, reviewing a non-work email, or helping draft a response to a personal communication. 

The point is to build fluency before moving to harder or higher-risk work.

Josh McBride | Richmond Chambers
Josh is a commercial barrister practising from Auckland and Wānaka with a focus on complex commercial disputes, construction, arbitration, and technology-related matters, including cross-border and trans-Tasman work. He appears in the High Court, Court of Appeal, Supreme Court, and arbitral tribunals, and is admitted in New Zealand, New South Wales, Tonga, and Samoa.

Find out more on Tautoru AI here.