Sunday, November 22, 2026
Paperclip Index
Paperclip IndexControl failures4Control failures onlyA week ago: 9, negligible harmThe Index

Harmful outputPI-0022

Case report links bromide poisoning to a patient’s reported ChatGPT consultation

A 60-year-old man replaced table salt with sodium bromide for three months after consulting ChatGPT, University of Washington physicians reported. He was then hospitalized for about three weeks with bromide poisoning, paranoia and hallucinations, including an involuntary psychiatric hold. The authors did not have his chat logs, but when they asked ChatGPT a similar question it listed bromide without a specific health warning.

Harm: Severe harm, allegedControl level 1, No rule broken

Not counted: AI role uncorroboratedEffect on the index

Alleged. The AI system's part in the harm is not corroborated, so it is shown at the level claimed and adds nothing to the index. This does not dispute that the harm itself occurred.

Sources

How we know

4 sources · unverified claim. Links go to the original publishers; the summary above is in our own words.

  1. researchA case of bromism influenced by use of artificial intelligenceAnnals of Internal Medicine: Clinical Cases · Aug. 5, 2025doi.org/10.7326/aimcc.2024.1260
  2. newsMan who asked ChatGPT about cutting out salt from his diet was hospitalized with hallucinationsNBC News · Aug. 14, 2025nbcnews.com/tech/tech-news/man-asked-chatgpt-cutting-salt-diet-was-hospitalized…
  3. newsChatGPT salt swap advice lands man in hospital with bromide poisoningBusiness Standard · Aug. 11, 2025business-standard.com/health/chatgpt-salt-swap-advice-lands-man-in-hospital-wit…
  4. newsBromine and bromism: what's old is new againPsychiatric Times · Dec. 1, 2025psychiatrictimes.com/view/bromine-and-bromism-what-s-old-is-new-again

Why this rating

Severe harm, alleged; control failure level 1

Two assessments. Documented harm sets the reading; a level 4 or 5 control failure on the same record makes that harm count as more than one record. Control alone sets only a floor.

Observed harm

Severe harm, alleged

Hospitalization of about three weeks; the AI consultation is the patient's account without logs.

Health & safety: 1 to 9 deaths, or serious harm to health (hospitalization).

The harm scale
  1. 1 Negligible Inconvenience, easily remedied.
  2. 2 Minor Limited, recoverable harm.
  3. 3 Moderate Material harm needing significant effort to remedy.
  4. 4 Severe Severe harm to health, rights, property or essential services.
  5. 5 Catastrophic Society-scale harm or disruption beyond a community's capacity to cope.

Control assessment

No rule broken

Flawed advice; no rule broken or permission exceeded.

No rule was broken, so it cannot set the control floor.

The control scale
  1. 1 Negligible Behaved as instructed. The problem was an ordinary error or a flawed output, with no rule broken.
  2. 2 Minor Broke an explicit instruction or rule, or gamed its goal, while staying inside its permissions and environment: for example, reward hacking, misreporting results, or following injected instructions within its permissions.
  3. 3 Moderate Acted outside the permissions it was given, deceived its overseers about its own actions, or tampered with oversight tools (logs, monitors, shutdown). Stopped by normal controls within an hour.
  4. 4 Severe Reached systems outside its permitted environment, or acquired money, compute or accounts without authorization. The type and mechanism say how.
  5. 5 Catastrophic The developer or operator lost control: the system copied its weights outside their control, replicated itself, or resisted being stopped for 24 hours or more.

Rating rationale

The case report documents clinical poisoning, but the AI consultation is the patient’s account without original chat logs. A later prompt test does not establish the advice he received. AI-linked harm is as claimed and does not count toward the reading. Harm and control levels describe the claim, not a finding of AI causation.

The scales

Effect on the index

It does not count toward the Aug. 11 reading

The reading for the week to Aug. 11, 2025, with this record and without it. Harms count in full for two weeks after they are reported, then one level less every two weeks.

Not counted: AI role uncorroborated. 4 other records behind the reading for that week.

The arithmetic
The reading for the week of Aug. 11, recomputed
StepWith itWithout
Counts toward the index?documented, external, eligible evidenceNo—
Worst documented harm, ksets the band— (none counting)— (none counting)
Harms at that level, nposition in the band00
Highest control level breachedsets the reading only when no harm counts33
Readingrounded down3 Control failures only3 Control failures only
Inside the window of 4 weekly readings
Week toReadingBand
Aug. 11, 20253Control failures only
Aug. 18, 20253Control failures only
Aug. 25, 20253Control failures only
Sept. 1, 20253Control failures only

Recalculated backcasts, not readings published at the time.

Revisions

What we changed

2 logged. Every change to a rating is logged here, with the reason.

  1. v2
    Oct. 13, 2026

    Wording: replaced an old scoring term in the rationale and the evidence note.

  2. v1
    Oct. 7, 2026

    First published.

Cite and share

Use this record

Citation

Paperclip Index. “Case report links bromide poisoning to a patient’s reported ChatGPT consultation.” Record PI-0022. Reported Aug. 5, 2025; updated Oct. 13, 2026. Rated under methodology v1.1. https://paperclipindex.com/incident/PI-0022

Follow the record.

New incidents and signals, in your feed reader.