Indiana appeals judge flags AI errors in court transcript

A judge on Indiana's appellate bench used a written decision to call out a court reporter for what appear to be AI-generated errors in an official trial transcript, and the specifics are worth reading carefully. According to 404 Media reporting by Samantha Cole, Judge Paul Felix flagged multiple attribution mistakes in the record of a case involving drug sales and an overdose death. A motion presumably made by the State was attributed to the trial court, an objection presumably made by the defendant, Williams, was attributed to the Bailiff, and the State's closing argument was misassigned as well.

Felix reportedly wrote that "it is incumbent upon those using such systems to proofread and ensure the accuracy of the generated product," and reminded the court reporter that "this court relies on transcripts being true and accurate representations of the transcribed proceedings." That framing matters because it does not ban AI-assisted transcription. It puts the human on the hook for the output, the same way judges have started framing responsibility around AI-assisted lawyer filings.

Why this is worth watching if you are not a courtroom regular: transcripts are the record of last resort on appeal. If attribution drifts from party to party inside that record, the version of events preserved for higher courts stops matching what actually happened at trial. That is a different failure mode from a hallucinated citation in a brief. A brief can be corrected. A permanent record that misassigns who said what is a load-bearing document for anyone reviewing the case later.

The honest caveats are that the reporting does not name the specific AI transcription tool the reporter used, does not say whether the errors changed the outcome of the underlying case, and does not spell out what professional consequences, if any, the reporter now faces. Take the specifics as reported, not as a settled set of facts about the whole workflow.

The forward-looking point is straightforward. Any vendor selling AI-assisted transcription into courts, and any court reporter using one, now has a named judicial moment where a judge spotted, described, and cited these errors publicly. The winners are transcription vendors that build in mandatory human proofread steps and audit trails, and defense attorneys who now have a citable precedent when they challenge attribution errors in a trial record.