AI in legal services: why I’m excited, and the benefits to consumers of regulation

The Legal Services Board has published two reports on AI in legal services. This blog shares some of what I found most striking while leading the research, including the voices of the people who took part.

I’ll be honest: I find AI genuinely exciting. The idea that someone facing a housing problem at 11pm on a Sunday, who can’t afford a solicitor and doesn’t know where to turn, could get clear and useful guidance from an AI tool feels like a real step forward. Around a third of adults in England and Wales with legal problems get no professional help at all. For many of them, AI could be transformative.

So, when we commissioned research, I approached it as someone who wanted to understand how we could make that potential a reality and ensure that regulation supports innovation while protecting people who need legal advice.

Consumers aren’t anti-technology. They’re anti-risk.

I expected to find anxiety about AI amongst consumers, scepticism, perhaps resistance. What I found instead was pragmatism.

People were broadly open to AI in legal services. They could see why it might help. As one participant put it: “Some people might not have the time, they’re busy working or they’ve got families. They might not have time to go and see a solicitor, but they can just ask it in the evening when they get home.”

That’s not a niche view. 70% of survey respondents expect AI to make legal services easier to use. Two-thirds expect it to improve accessibility. Nearly as many expect it to become more affordable.

But that openness came with a clear condition. People were unwilling to trade away basic protections for convenience. They identified five things they considered non-negotiable, regardless of price point or platform.

1. They want accuracy they can rely on, with clear quality checks so advice isn’t misleading or unsafe.
2. They expect to be asked before AI takes any consequential action on their behalf.
3. They want a human available when the stakes are high.
4. They expect a fair complaints process and a route to compensation if something goes wrong.
5. They expect their personal data to be protected and used transparently.

These aren’t unreasonable demands. They are, in most regulated contexts, baseline expectations. What the research found, though, is that many AI legal tools do not meet these expectations.

The gaps we found

In parallel with the consumer research, we reviewed some of the standards that currently govern AI-powered legal tools for consumers. The picture was striking.

Standards specifically designed for consumer-facing legal AI are almost entirely absent. Most of what exists was created for general AI governance, data protection, or professional conduct for lawyers. It was never designed for tools that directly communicate with members of the public about their legal rights.

Of the standards we reviewed, 58% are non-binding guidance with no enforcement mechanisms. That figure stayed with me. They represent good intentions but without guaranteed accountability.

The weakest areas were precisely the ones consumers cared most about: identifying vulnerable users, directing people to regulated support when they need it, and providing a clear route to redress when things go wrong.

One interviewee captured the problem plainly: “any general AI tool should have to tell people upfront: I’m not a lawyer.” That principle is sound. The challenge is that right now, there is no requirement for it.

Regulated lawyers have a safety net. Direct AI users often don’t

What the research exposed isn’t just a gap in standards. It’s a structural problem.

When a consumer uses AI through a regulated solicitor, professional obligations apply. There are complaint routes, oversight, and indemnity insurance. If something goes wrong, there are places to go.

When a consumer uses a direct-to-consumer AI legal tool, those protections often don’t exist. General consumer and data protection law applies in theory, but it wasn’t designed for AI-specific risks like hallucination or for the particular vulnerabilities of someone trying to understand their housing rights at midnight with no professional to turn to.

As one consumer representative told us: “People are going to ChatGPT because they don’t know where else to go.”

That observation stays with me. Because the people most likely to use these tools without professional support are often the same people who are most at risk if the tool gets it wrong, and least able to challenge the outcome.

Why this matters now

The tools are already out there. People are already using them. Clients are arriving at advice services like Citizens Advice having already consulted ChatGPT, with results of variable quality. The question isn’t whether this is happening. It’s whether we want it to happen without any baseline protections in place.

I want to be clear: the answer is not to slow down innovation. The goal is to make the potential real, not to obstruct it. An AI tool that helps someone understand their options in a housing dispute, or navigate a debt problem, or know their rights at work, is genuinely valuable. For people who would otherwise get no help at all, it could be the difference between resolving a problem and letting it spiral.

But “first do no harm” should apply here too. One interviewee proposed thinking about consumer AI legal tools like a domestic first aid kit: accessible guidance for common situations, with clear signposting to professional help when things go beyond what first aid can manage. That framing feels right to me. The goal isn’t to replace professional services. It’s to provide a trustworthy first layer, with clear routes to something more when needed.

What the LSB is doing next

The LSB is sharing these findings because they matter for the whole sector as we innovate and embrace technology. We’re not the only ones who need to act on them.

We will focus our efforts where regulation can have the greatest impact, working with regulators, government, innovators, and consumer organisations to define what targeted and proportionate action should look like. We’re also participating in the government’s AI Growth Lab, supporting lawtech firms to test products responsibly and helping identify where the current framework creates unnecessary barriers.

The research shows that consumers are ready for AI in legal services.  The people who took part in this research were not asking for perfection. They were asking for honesty, accountability, and somewhere to turn if things go wrong. Is that really too much to ask of any service that helps people with some of the most stressful situations in their lives?

Both reports are available to download. We welcome responses from regulators, innovators, consumer organisations, and anyone working in this space.