AI in Justice Must Be People-Centred — Not Just Another Tool to Process Vulnerable People Faster

In this article

• Why AI is now a serious access to justice issue, not a future theory.
• What AI can genuinely help with in courts and legal support.
• Why family justice needs a different conversation from commercial legal tech.
• The risks: hallucinations, bias, digital exclusion, deepfakes, privacy and over-reliance.
• What safe, people-centred AI in the family justice system should look like.
• Practical guidance for litigants in person using AI tools.

AI is already in the justice system

The debate about artificial intelligence and justice is no longer theoretical. Courts, lawyers, public bodies, government departments, judges, charities and litigants in person are already using AI in different ways. Some use it openly. Some use it cautiously. Some use it badly. Some use it without really understanding what it is doing.

That matters because justice is not an ordinary administrative service. A mistake in an online shopping recommendation is irritating. A mistake in a court process can affect liberty, housing, money, children, safety, contact arrangements, immigration status, credibility and protection from abuse.

In family proceedings, the stakes are even sharper. The court is often dealing with frightened parents, children whose voices are filtered through adults, allegations of domestic abuse, coercive control, parental alienation arguments, safeguarding concerns, mental health issues, poverty, trauma, language barriers and litigants in person who are expected to navigate complex legal procedure with little or no help.

So yes, AI can help. But the question is not simply:

“Can AI make the justice system faster?”

The better question is:

“Can AI make the justice system fairer, safer, more understandable and more accountable for the people who actually have to live with its decisions?”

The access to justice opportunity is real

I do not take an anti-technology position. That would be unrealistic and, frankly, unhelpful. Used properly, AI could support access to justice in ways that are genuinely meaningful.

It can help people turn messy information into a chronology. It can help summarise long documents. It can help translate legal jargon into plain English. It can help identify missing evidence. It can help a litigant in person understand what a direction means, what a bundle is, what a witness statement should contain and why a court needs evidence rather than emotion.

It can also help professionals. Properly governed AI could assist with transcription, document search, issue spotting, safeguarding triage, case management, resource allocation and identifying repeated procedural bottlenecks. Those things matter because delay is not neutral. In family cases, delay can become a safeguarding issue. A child waiting months for clarity is not living in a holding pattern. They are growing up inside uncertainty.

This is why I am interested in AI. Not because it is fashionable. Not because it allows legal services to look innovative. But because it could help real people who are currently drowning in documents, fear, deadlines and legal language they were never taught to understand.

But speed is not the same as justice

The danger is that AI gets sold to the justice system as an efficiency machine. Faster triage. Faster drafting. Faster listing. Faster case progression. Faster disposal.

There is nothing wrong with speed where speed serves justice. But speed can also become dangerous when the system is already struggling to hear people properly.

A survivor of coercive control does not simply need a faster process. They need a process that understands patterns of behaviour. A child does not simply need a faster decision. They need a safe, welfare-focused decision based on proper evidence. A litigant in person does not simply need an online portal. They need to understand what is being asked of them, why it matters and how to respond without accidentally harming their own case.

That is why AI in justice must be judged by more than productivity. We need to ask whether it improves fairness, understanding, safeguarding, accountability, participation and trust.

The family court needs a trauma-informed AI conversation

Much of the legal technology conversation is still shaped by commercial law: contract review, disclosure, due diligence, legal research and productivity. Those use cases matter, but family justice is different.

Family proceedings are not simply about processing legal disputes. They involve relationships, children, fear, housing, money, safety, care arrangements, allegations of abuse and power imbalance. A person involved in family proceedings may be sleep-deprived, frightened, financially controlled, isolated, threatened, neurodivergent, traumatised or trying to protect a child while being accused of obstructing contact.

That means AI tools designed for family justice must be trauma-informed and safeguarding-aware from the outset. They must not treat all conflict as mutual. They must not flatten coercive control into “poor communication”. They must not encourage unsafe negotiation. They must not push survivors into mediation-style language where there are unresolved risk issues. They must not confuse calm presentation with truth, or emotional distress with unreliability.

AI must not be trained only to tidy documents. It must be designed to preserve context.

The hallucination problem is not a minor technical issue

One of the most serious risks with generative AI is hallucination. That means the tool may produce information that sounds confident but is false. In law, this can include invented cases, wrong legal tests, out-of-date procedure, fake citations or a misleading summary of what the law says.

This is not an abstract risk. Courts have already had to deal with legal material containing fake authorities. The lesson is blunt: fake cases are not “cosmetic errors”. They can mislead the court, waste time, damage credibility and create professional consequences.

For litigants in person, the risk is slightly different but equally serious. A person may use AI because they cannot afford a lawyer. The AI may produce a document that looks impressive. It may use legal language. It may cite principles confidently. But if the law is wrong, the cases are false or the facts are overstated, that person may walk into court with a document that harms them.

Confidentiality and privacy are major safeguarding issues

Family cases often involve intensely private information: children’s names, addresses, schools, medical records, allegations of abuse, police material, Cafcass reports, social care information, financial disclosure, intimate messages and evidence that may be subject to court restrictions.

People must be extremely careful about putting this material into public AI tools. A parent should not upload a full bundle, a Cafcass report, a child’s medical information, private messages or identifying safeguarding material into a public chatbot without understanding where that data goes, how it may be stored and whether it could be used for training or reviewed by others.

This is not just a data protection point. It is a safety point. In domestic abuse and coercive control cases, information leakage can create real-world risk.

Bias does not disappear because the decision is digital

AI systems learn from data. If the data reflects bias, gaps, stereotypes or historic unfairness, the output may reproduce those problems. That matters profoundly in family justice.

We already know that survivors of domestic abuse are often disbelieved, minimised or expected to present in a particular way. We already know that mothers can be scrutinised for their reactions while the perpetrator’s behaviour is contextualised. We already know that fathers who are safe and child-focused can also be unfairly stereotyped. We already know that poverty, disability, race, language, trauma and neurodivergence can affect how people are perceived.

AI does not automatically solve that. It may make it worse if it is trained on historic material that contains the same assumptions.

A tool that summarises a domestic abuse statement, for example, might remove the very details that show the pattern. It might reduce years of coercive control to “relationship conflict”. It might prioritise incidents with police evidence while minimising financial control, isolation, intimidation, threats, sexual coercion or litigation abuse.

Digital exclusion is an access to justice problem

There is a real risk that AI improves the experience of people who are already digitally confident, while leaving others even further behind.

Not everyone has a laptop, stable internet, quiet space, digital literacy, English fluency, confidence with forms, or the ability to pay for premium tools. Some people only have a phone. Some are being monitored by an abusive partner or ex-partner. Some cannot safely store documents. Some cannot receive email without risk. Some have disabilities or literacy difficulties. Some are trying to manage court deadlines while working, caring for children or escaping abuse.

A justice system that assumes everyone can use digital tools safely and confidently is not accessible. It is selective.

That is why AI must be accompanied by human support, clear guidance, offline alternatives, accessible design and proper safeguards. Otherwise we simply create a two-tier system: those who can use AI well, and those who are processed by systems they do not understand.

Deepfakes and AI-generated abuse evidence are coming into family cases

The family court also needs to prepare for AI-generated harm. Deepfake images, fake audio, manipulated screenshots, synthetic messages and AI-generated impersonation are not science fiction. They are already part of the wider abuse landscape.

For survivors of domestic abuse, this matters because technology is often used as a tool of control. AI can be used to humiliate, threaten, stalk, impersonate, discredit or overwhelm. It can also be used to create confusion around evidence: “that message is fake”, “that recording is AI”, “that image was manipulated”.

Courts will need better literacy around digital evidence. Parties will need clearer ways to preserve metadata, source material and context. Professionals will need to understand both sides of the problem: AI-generated abuse and false allegations that genuine evidence is AI-generated.

What safe AI in justice should look like

I am not calling for AI to be kept out of justice. I am calling for it to be used properly.

In my view, any AI system used in justice — particularly family justice — should meet basic safeguards before it is trusted anywhere near decisions affecting children, safety, liberty, homes or fundamental rights.

A practical guide for litigants in person using AI

Many litigants in person are already using AI because they cannot afford legal representation. I understand why. Used carefully, it can be a useful support tool. Used carelessly, it can damage a case.

The safest approach is to use AI as an assistant, not an authority. It can help you organise your thinking. It cannot replace evidence, legal advice, court rules or careful strategy.

What this means for JSH Law

My work is already evidence-led. I help people turn confusion into structure: chronologies, schedules, statements, position statements, appeal documents, hearing preparation and clearer case analysis.

AI fits into that work only if it strengthens the human part of the process. The aim is not to automate people’s pain. The aim is to help people present their case more clearly, safely and effectively.

That means using technology to make the court’s job easier without stripping out the facts that matter. It means helping litigants in person avoid avoidable mistakes. It means making documents more focused, not more performative. It means recognising patterns of coercive control, not reducing everything to “communication difficulties”. It means preparing evidence in a way that is calm, structured and credible.

The future: AI should help people tell their story once, properly

One of the most powerful possibilities for AI in family justice is not flashy at all. It is the possibility of reducing repetition.

Parents and children often have to tell the same painful story repeatedly: to police, solicitors, Cafcass, social workers, domestic abuse services, schools, doctors, mediators, judges and support workers. Each retelling can be exhausting. Each retelling can lose detail. Each retelling can be misunderstood.

A properly designed, secure and trauma-informed system could help people preserve their account, organise it safely, update it when necessary and make sure the right information reaches the right person at the right time.

That would be technology serving people.

But a badly designed system could do the opposite. It could compress their story into tick boxes. It could miss coercive control. It could treat distress as unreliability. It could push unsafe settlement. It could make decisions feel even more remote and unchallengeable.

That would be people serving technology.

The justice system must choose the first path.

Conclusion

AI is not the enemy of justice. But neither is it a magic fix for years of underfunding, delay, legal aid gaps, poor safeguarding practice, inaccessible procedure or the burden placed on litigants in person.

The real test is whether AI helps people understand the system, participate fairly, protect children, present evidence properly and challenge decisions that affect their lives.

In family justice, the measure of success should not be how quickly a case is processed. It should be whether the process is safe, fair, evidence-led and child-focused.

AI should not make vulnerable people easier to process. It should make justice harder to avoid.

Need help preparing family court documents?

JSH Law helps litigants in person with evidence organisation, chronologies, schedules of allegations, witness statements, position statements, Cafcass response work, appeal paperwork, non-molestation order preparation and hearing support.

The aim is simple: clear documents, safer strategy, stronger evidence and a case that is easier for the court to understand.

Book a consultation Contact JSH Law