Sydney, In the Federal Court of Australia, applications by self-represented litigants, many using artificial intelligence to research and draft their claims, have skyrocketed.

The Fair Work Commission, a federal tribunal, has already seen a jump in filings, many of them "AI slop".
These experiences are also reflected in our centre's global database of AI-related cases. It contains 295 Australian cases in which a court or tribunal identified problematic use of AI in legal proceedings.
Examples include citing made-up case law, citing law that is real but irrelevant to the case, and using AI to write witness statements or submissions that aren't entirely accurate.
The database shows a huge increase in cases where these types of AI misuse has been mentioned over the past year, with self-represented litigants accounting for 81 per cent of Australian misuse cases.
Chatbots can make poor lawyers
-
Access to courts is a fundamental part of access to justice. But starting court proceedings without a viable or valid claim wastes time, causes stress and may have severe financial consequences for the person bringing the case.
People might use chatbots built on large language models to help them with a legal matter in different ways: 1. Advice about the law generally but also how it specifically applies to their own situation
{{/usCountry}}People might use chatbots built on large language models to help them with a legal matter in different ways: 1. Advice about the law generally but also how it specifically applies to their own situation
{{/usCountry}}2. Drafting of court documents to commence or defend a case
3. Drafting of court documents to support the case, such as submissions or statements about the facts .
While AI might usefully assist people to do all of these things, generative AI chatbots are much more sycophantic than humans.
Lawyers have a duty to always act in their clients' best interests, which sometimes means challenging clients' views and delivering unpopular opinions.
AI chatbots have no such obligations.
There may also be errors in the material that they generate. For example, in Sadnan Khan v Aldi Pty Ltd, the Fair Work Commission took the unusual step of ordering costs of AUD 1,230 against a former Aldi employee who tried to sue for unfair dismissal. The commission found Khan had relied on AI-generated arguments even though he should have realised his claim was doomed to fail. .
To use AI successfully, people need to be able to check sources, verify information and ultimately explain the relevance of the legal arguments they are trying to make.
The higher the court, the riskier AI-use becomes
-
The main impact of poor use of AI is twofold. First, most people are not successful, either because their legal argument doesn't work, or the facts don't support the argument, or both.
Two unpublished papers using data from US courts in civil matters show a similar pattern: large increases in court filings by self-represented people using AI but no apparent change to win rates, which remain low.
Second, the recent Australian Federal Court data show unsuccessful claims are now working their way up the court food chain. A high proportion of the increase in filings is made up of appeals from people who were unsuccessful in lower courts or tribunals.
Litigating in higher courts, such as the Federal Court, is riskier. The "loser pays" rule of legal costs often applies . This is different from lower-tier courts and tribunals, which are often "no cost" jurisdictions, meaning costs are not routinely ordered.
Besides ordering costs, what have courts and tribunals been doing about this increasing workload?
As we explain in a recent article, Australian courts are not alone – similar issues are being faced in many countries.
Broadly, courts have adopted a pyramid of responses, beginning with education at the base. Most Australian courts have now released guidance about the responsible use of generative AI in legal proceedings.
The Fair Work Commission's guidance, effective from October 20 2026, explains the importance of real legal sources, and even suggests how to prompt AI chatbots more effectively.
In Singapore, courts are exploring their own AI-powered tools for users in small claims matters.
The next layer involves corrective measures, such as permitting a person to amend and refile documents, and giving warnings about future sanctions. Where stronger actions are warranted, courts and tribunals are responding with a range of deterrent options: striking out all or part of a claim or defence; imposing restrictions on filing of further materials; and of course the significant costs that can be incurred. At the very apex of the pyramid is the possibility of legal or even criminal sanction via the laws of contempt of court.
However, warnings and sanctions alone are unlikely to solve the problem of self-represented litigants misusing AI. The difficulty? People like the responses produced by generative AI. It feels supportive and encouraging, and can produce fluent-sounding arguments far quicker than a human could draft them.
Where access to legal advice can be costly and hard to obtain, AI chatbots are always available and often free. AI is already becoming part of how people engage with the justice system. The real challenge for courts is ensuring people have the knowledge and support needed to use AI responsibly to improve their case, instead of sabotaging it.
GSP
This article was generated from an automated news agency feed without modifications to text.