AI in the courts: What happens when efficiency leads to a collapse of the judicial system?

A lawyer filed nearly 38,000 court documents in just two days using an artificial intelligence tool. The incident, which took place in Chile’s civil courts, prompted the Santiago Committee of Civil Judges to ask the Supreme Court to take urgent measures, including a possible ban on the use of AI for this type of procedure. This episode is not an isolated case: it is one of a series of recent incidents, both in Chile and abroad, which demonstrate how artificial intelligence is already pushing the boundaries of the judicial system, long before any regulations capable of reining it in are in place. 

The case that set off the alarms

According to an official letter sent to the President of the Supreme Court, a lawyer used an ‘external agent’ powered by artificial intelligence to submit a massive number of applications to remove cases from the docket and to withdraw from representation and power of attorney in civil courts across the country. Between Saturday 25 and Monday 27 July, the system recorded 38,477 submissions, equivalent to between 500 and 900 documents per court – a figure which, according to the official letter itself, was still rising at the time of reporting.

The Santiago Committee of Civil Judges described the situation as an “abusive use of technological tools”, which could constitute a case of procedural abuse and breach the terms of use of the Virtual Judicial Office. The institutional response was drastic: requesting that the lawyer’s IP address be blocked, assessing potential disciplinary and criminal liability, and even banning the use of artificial intelligence for the submission of pleadings until such time as an institutional policy regulating its use is in place.

A recurring pattern

The Chilean case does not occur in a vacuum. Months earlier, a labour court in Brazil detected that a claim contained invisible text (written in white on a white background) directed at an AI system, seeking to alter how it processed the document. This was described as an act undermining the dignity of justice and resulted in sanctions against the lawyers responsible.

Both incidents, although different in their mechanics, share the same underlying issue: AI is being used in legal proceedings without any controls specifically designed to prevent its misuse. In one case, the content was manipulated to deceive the system; in the other, its automation capabilities were exploited to overwhelm it. In both instances, the consequence was the same: the courts had to react after the event, not before.

Why this matters

The promise of artificial intelligence in the legal sphere has always been associated with efficiency: less time, lower costs, greater management capacity. But these cases reveal the other side of that promise. When automation is used without limits or supervision, what is efficiency for one can become the operational collapse of an entire system. The committee’s own judges were explicit: ‘we do not have the technical or human resources available to deal with this contingency’.

Furthermore, the problem is not limited to workload. It raises fundamental questions about professional responsibility: to what extent can a lawyer delegate the exercise of their representation to an automated agent? What standard of care is required when the tool acts ‘on their behalf’ on a scale that no human being could match? These are questions that the Chilean disciplinary system is only just beginning to ask itself.

The regulatory vacuum

As these cases unfold, the regulatory gap is becoming increasingly apparent. Since 2 August, the European Union has required that content generated or modified using AI be identified, within the framework of its Artificial Intelligence Act. Spain, for its part, is making progress on a specific draft bill on algorithmic governance, aligned with the European AI Regulation.

Chile, by contrast, still does not have a comprehensive law on artificial intelligence in force. The bill addressing these uses has been pending since 2024. There is, however, a National Artificial Intelligence Policy and significant progress has been made on the new Personal Data Protection Act, which reinforces security obligations and the duty of care. However, none of these tools was designed with scenarios such as that of the 38,000 legal documents in mind: the use of automated agents within judicial platforms remains, in practice, a field without clear rules.

The case of the 38,000 legal documents is not merely a technological anecdote: it is a warning of what happens when the capacity for automation exceeds the institutional capacity to oversee it. Just as judicial prompt injection in Brazil showed that AI can be susceptible to manipulation, this Chilean case demonstrates that it can also be the vehicle for large-scale abuse, even without any obvious malicious intent.

The question that remains open for the judicial system and for legal practitioners is no longer whether artificial intelligence will feature in legal proceedings. It already does. The question is whether the institutions (and the lawyers themselves) will define the limits of its use before the next incident forces them to do so as a matter of urgency.