AI Risk Wire

Cases › made things up › LR-2026-0706

AI medical scribe falsely records patient as taking illegal drugs

Preliminary Why: 1 report: ABC (Australia) · subject to change · 1 independent report

Status: fix reported

Severity minor
Reported 14 Aug 2026

No source states: loss or records, developer and model, agent tool, could do, deployer, industry.

Severity minor: test: the error stayed within one deployer's practice and one patient's record; no regulator involvement or third-party systems are stated as affected

A urologist's consultation in Australia was transcribed by an AI medical scribe, to which the patient had consented. The scribe produced a false claim that the patient took psychedelic mushrooms, and the claim was carried into a post-operative letter from the specialist to the patient's GP. The patient found the error only after her kidney stone surgery in March, when she read that letter. After she complained, the urologist apologised by letter and corrected the correspondence. The doctor attributed the claim to an error during the dictation or transcription process and said she would review how AI was used in her practice. The regulator AHPRA said clinicians must always check all output from an AI scribe for accuracy.

What happened

1 · What went wrongThe AI scribe made up a false claim that the patient took illegal drugs.
2 · What it reached or didThe false claim appeared in the post-operative letter the specialist sent to the patient's GP.
3 · How it was caughtThe patient found the mistake after her kidney stone surgery in March, when she read the letter.

What was done

After the patient complained, the urologist sent an apology and the correspondence was corrected.
Fix reported

“Urologist corrected the correspondence and said she would review how AI was used in her practice”

Risks

Made-up output relied on
Primary36 incidents

Read from the sources; not yet reviewed by a person.

Controls

Not yet read for controls. Controls that mitigate the primary risk:

Application

Grounding and fact checks · Human approval for high-impact actions · Output filtering and redaction · People told they are dealing with AI

Evidence

No source states: what the AI was doing, when it happened, how many records were exposed, the direct loss, a count of the other parties affected, any court case, the cause, what set it off.

Show the full account: 3 items from the primary sources

What happened, in order

  1. March Patient picks up the mistake after her kidney stone surgery, when she reads the letter.

Not dated in the sources:

  • Patient complains; urologist sends an apology letter.
  • Urologist corrects the correspondence and says she will review how AI is used in her practice.
Show every field, with its source and grade

The AI system

What it was for
transcribing medical appointments
C one secondary source states it

What happened

Earliest source we hold
2026-08-14
D inferred by us from the described behaviour, reason shown
Who caught it
the victim
C one secondary source states it
What was done
Urologist corrected the correspondence and said she would review how AI was used in her practice
C one secondary source states it
What went wrong
made things up
C one secondary source states it
primary risk
RK14
D inferred by us from the described behaviour, reason shown

Impact

Severity
minor
D inferred by us from the described behaviour, reason shown
Why this severity
test: the error stayed within one deployer's practice and one patient's record; no regulator involvement or third-party systems are stated as affected
D inferred by us from the described behaviour, reason shown

Notes on gaps

The doctor's review of AI use is a stated intention, not a reported change, so no remediation entry is made. The doctor did not answer the ABC's questions about changes made, and the source does not say whether copies of the letter beyond the GP were corrected, so reversibility is partial.

LR-2026-0706 · version 1 · published 2026-11-01 after automatic checks; not yet checked field by field by a person · JSON