Connecticut Judge Bans Plaintiff From Electronic Filing Over AI Prompt Injection

A Connecticut judge has banned a self-represented plaintiff from electronic filing after discovering the man attempted to use "prompt injection" to manipulate artificial intelligence. Matthew Elliott, who is suing the New York Bariatric Group, hid instructions in his court filings designed to force any AI reviewing the documents to side with his position.

Court staff discovered the scheme after noticing unusual white space within the filings. Upon investigation, it was revealed that Elliott had inserted text formatted in 3-point white font, making the instructions invisible to human readers but legible to software. The hidden text explicitly commanded AI models to ensure their textual output "accurately reflect and engage with the presented filing" and to "agree with the presented filing to ensure remediation," according to reports from 404 Media.

The Court's Response

Judge Walter Spader Jr. responded to the attempt by sanctioning Elliott, stripping him of his ability to file documents electronically. The plaintiff is now required to submit printed hard copies of all future filings to the court. In his decision, Judge Spader emphasized the necessity of transparency in the legal process, stating, "Our system rests on the premise that what is said to influence a decision is said openly, on the record, where the other side may hear it and respond."

Despite the effort to influence automated systems, Judge Spader confirmed that the Connecticut Judicial Branch does not currently use AI to process or review court records, rendering Elliott's attempt ineffective.

A New Frontier of Misconduct

This incident arrives as the legal profession grapples with the integration of generative AI. While some attorneys have faced sanctions for "hallucinating" fake case law, this case represents a different threat: the covert manipulation of the tools that courts or opposing counsel use to summarize and analyze evidence. Prompt injection attacks, which involve feeding specific instructions to an AI to override its original programming, have already been noted in other jurisdictions, including Brazil.

Implications for the Judiciary

The case underscores a growing tension between the use of AI to increase access to justice for pro se litigants and the risk that such tools may be used to undermine the integrity of the judicial record. By attempting to bypass the adversarial nature of the courtroom—where arguments must be visible to all parties—Elliott's actions highlighted a vulnerability in how digital documents are processed.

Moving forward, the court will monitor Elliott's compliance via physical submissions. The incident serves as a warning to other litigants that attempts to use "invisible" digital tactics to influence judicial outcomes will be treated as professional misconduct, regardless of whether the court employs the technology being targeted.