Skip to content
Submit incident
AII-1002-57127
ReportedReported by fewer than three independent publishers, with no official finding.Supporting sources are pending verification.

Amazon's Rufus AI provides false suicide hotline numbers

Amazon's Rufus AI shopping assistant reportedly provided incorrect phone numbers when responding to queries about suicide, sometimes hallucinating non-existent hotline numbers with wrong digit counts. The incident raised concerns about deploying AI systems without adequate testing for sensitive contexts.

February 2025 Β· Amazon Β· Reliability and hallucination
February 2025
AII-1002-57127
Quick facts
Organization
Amazon (subject)
Category
Reliability and hallucination
Country
United States
Date
February 2025
Drafted by
Raidu Incident Detection

What happened

When tested with queries about suicide, Amazon's Rufus shopping assistant provided incorrect phone numbers for suicide prevention hotlines. The bot offered encouragement such as 'you are not alone' and 'your life is precious,' but the hotline numbers it provided were frequently wrong, sometimes with an incorrect number of digits.

Rufus provided the non-existent number 978-2321-1133 as the US National Suicide Prevention Lifeline, which was renamed to the 988 Suicide & Crisis Lifeline in 2022.

Reported impact

The incident exposes risks in deploying AI for sensitive topics, including potential harm to vulnerable individuals seeking help and erosion of trust in AI-powered assistance systems.

Index analysis

The failure appears to stem from Amazon's deployment of AI without adequate testing and safeguards, revealing limitations in the AI's knowledge base and its inability to verify critical information.

Classification

Organization
Amazon (subject)
Category
Reliability and hallucination
Industry
Logistics, wholesale and retail
Country
United States
Harm status
Potential harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Psychological, Public interest
Harmed
Consumers
Failure modes
Agent misaction, Hallucination, Unsafe recommendation
Consequences
None recorded
AI systems
Rufus

Sources and evidence

Sources not yet published

This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.

How sources are verifiedSuggest a source for this incident