AI for people without a lawyer: what help is safe and what is not
By the ChamberOne team 4 minute read
A woman facing an eviction notice has no lawyer and a deadline she does not understand. At home that evening she asks an AI tool what to do. The answer is fluent and confident. It may be right. It may describe the law of another country, miss a deadline or invent a case she then quotes to a judge. For legal aid workers and court-user support staff, this is already happening in the queue in front of them. This article separates the kinds of help where AI is reasonably safe from those where it is not, and offers ways to point people towards proper help.
Key points
- Explaining forms, organising facts and preparing questions are lower-risk uses.
- Strategy, deadlines and anything going to a court need checking by someone who can check it.
- Tell users plainly that answers may be wrong or about another country, and show a route to a person.
- Keep a current signposting list and warn people about what they type into a tool.
Why the stakes are different
A lawyer who uses AI checks the output with training and experience. A person without a lawyer cannot. They cannot tell a sound answer from a fluent wrong one, and they may act on it with real consequences: a missed deadline, a form filled in wrongly, a position conceded, a home lost.
The reasons people go without a lawyer also matter. Cost, distance and fear are common, and these are people who may have no one else to ask. Honest guidance has to account for that, and should not simply say 'get a lawyer' where one is not available.
Help that is generally lower risk
- Explaining what a form or notice is asking for, in plain words.
- Turning a jumble of facts into a clear timeline the person can take to an adviser.
- Explaining what a legal term means in general.
- Preparing a list of questions to ask at an advice session or a court counter.
- Translating or simplifying a letter, with a check by a person where it matters.
- Helping someone say in their own words what happened, so they can give it to an adviser.
Help that carries real risk
The risk rises as the task moves from explaining to deciding. Whether to settle, whether to defend, which claim to bring, what to say in a hearing and whether a deadline has passed are decisions that turn on the law of a place and the facts of a case. A tool may not know either.
Particular danger lies in deadlines and limitation periods. A wrong date can end a claim. Equally dangerous is the made-up authority. People without lawyers have filed documents citing cases that do not exist, and courts have noticed. Anything that is going to a court or an opposing party needs checking by someone who can check it.
- Explaining what a form or notice asks for
- Turning facts into a clear timeline
- Preparing questions for an adviser
- Explaining a legal term in general
- Whether to settle or defend
- Which claim to bring
- What to say in a hearing
- Whether a deadline has passed
What a service should tell users
If your service offers or recommends AI tools, say plainly what they can and cannot do. Say that answers may be wrong, that they may be about a different country, that deadlines must be confirmed with the court or a legal adviser, and that nothing the tool says is legal advice. Put this in plain language, and in the languages your users speak.
Show a route to a person. A phone number, an appointment link or the address of the nearest advice centre should appear next to any tool, not on a distant page.
Signposting in practice
Build a short list of places that can help: legal aid offices, law centres, pro bono clinics, court help desks, bar association referral services and, where they exist, duty advisers. Keep it current, and check each entry on a schedule. Dead links and closed offices harm people who have little time.
When someone brings you AI-generated text for a court filing, do not simply refuse it. Read it with them. Ask where each statement came from, whether they have seen the case or statute, and what a court would make of it. That conversation teaches more than a warning.
Confidentiality and the user's own data
People tend to type everything into a chat window, including names, addresses, medical details and immigration status. Warn them that what they enter may be stored or seen by others. Suggest that they leave out identifying details when asking general questions, and that they not share the text with others if it describes their case.
Staff who use tools to help users carry the same duty of care. Do not paste a user's personal story into a general-purpose tool without their agreement and without checking where it goes.
A short scenario
Suppose a court help desk supports people filling in a response to a claim. A volunteer sits with a user and uses an approved tool to explain each box in plain language. The tool does not suggest what the person should say about the dispute. When the user asks, 'should I admit this?', the volunteer says that is a question for an adviser, writes it on a list of questions, and gives the person the number of the nearest legal aid office. The help was real, and it stopped where the risk began.
Special cases that need a person
- Anything involving a threat to safety, such as domestic violence or an immediate risk of homelessness.
- Immigration and asylum questions, where an error can have severe and lasting effects.
- Criminal charges, where the person may be at risk of custody.
- Documents that will be signed under oath or filed with a court.
- Any matter where a limitation period or appeal deadline may be close.
Design principles for a service
Keep the tool to a narrow job. Limit it to explaining forms, organising facts and preparing questions. Tell it to decline strategy and deadline advice, and test that it does. Review real conversations, with the users' permission and with identifying detail removed, to see where it goes wrong. Involve lawyers in writing and checking the instructions. And measure what matters: whether people reach proper help, not how many chats took place.
- Keep the tool to explaining forms, organising facts and preparing questions
- Test that it declines strategy and deadline advice
- Involve lawyers in writing and checking the instructions
- Let it advise on strategy or deadlines
- Measure success by how many chats took place
This article is general information about the law, legal practice and technology, current when it was published. It is not legal advice. Laws and guidance change: check the sources linked above and take advice on your own circumstances from a lawyer qualified where you are.
Read next
AI and access to justice: an honest assessment
Where AI can lower the cost of routine legal help, where it cannot replace advice, and who bears the risk when it is wrong.
5 minute readClient portals: what self-service does for clients and the firm
Which questions a portal answers, what to keep out of it, and how it reduces routine calls.
5 minute readAI chatbots on law firm websites: benefits, risks and rules
A chatbot can help with intake, but it can also say too much. Decide its limits before it goes live.
4 minute read