WHO S.A.R.A.H. AI health bot provides inaccurate medical information
The WHO's AI-powered health advisor, S.A.R.A.H., was reported to provide inconsistent and incorrect medical information. Despite its status as a research prototype, the bot was accessible to the public and frequently hallucinated facts.
S.A.R.A.H. doesn't have up-to-date medical data and can hallucinate.
Key facts
- What
- The WHO's AI-powered health advisor, S.A.R.A.H., was reported to provide inconsistent and incorrect medical information.
- Incident date
- Apr 18, 2024
- Who
- World Health Organization
- Failure mode
- Hallucination
- AI surface
- Chatbot
- Severity
- Medium
What happened
The WHO released S.A.R.A.H. as a prototype health assistant to explore the use of AI in public health. In April 2024, reports emerged that the bot provided wrong answers to medical queries and hallucinated information. The system often defaulted to generic advice to consult professionals when it could not provide a reliable answer.
What broke inside the model
- 01 · TriggerA user asks for a fact, a citation, or a figure.
- 02 · Model stepThe model writes a fluent, confident answer.
- 03 · Control gapNothing ties the claim back to a real source.
- 04 · FailureA fabricated fact ships as if it were verified.
- 05 · ConsequenceThe false claim reaches a customer, a court, or the public.
Confidence holds, and even spikes, as the claim detaches from any source.
The bot lacked access to real-time, up-to-date medical data and suffered from LLM hallucinations. These issues resulted in inconsistent responses and factual inaccuracies when handling complex health-related questions.
What it cost
Sources
- PressAI-Powered World Health Chatbot Is Flubbing Some Answersbloomberg.com
- PressWHO's new AI-powered chatbot SARAH is giving wrong medical answershealth.economictimes.indiatimes.com
Cite this entry
https://failureindex.ai/failures/health-bot-provides-inaccurate-medical-informationAI Failure Index. "WHO S.A.R.A.H. AI health bot provides inaccurate medical information" (FI-0636). Realm Labs. https://failureindex.ai/failures/health-bot-provides-inaccurate-medical-information (indexed Jun 22, 2026).Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0636. Full dataset at /data.
Note from Realm Labs, the Index steward
How Realm would have caught this
- Prism
- OmniGuard
- AI Detection & Response (AIDR)
A runtime layer that watches the model's internal state can flag the moment a model commits to a claim it has no support for, and hold or reroute the response before it reaches a user. Realm reads those signals in real time rather than grading the transcript after the fact.