Amazon Q chatbot allegedly leaks confidential AWS data and hallucinations
Amazon's AI chatbot, Q, allegedly suffered from severe hallucinations and leaked confidential company data, including data center locations. While internal documents flagged the issue as a significant incident, Amazon spokespeople denied that any confidential information was leaked.
Amazon Q is experiencing severe hallucinations and leaking confidential data.
Key facts
- What
- Amazon's AI chatbot, Q, allegedly suffered from severe hallucinations and leaked confidential company data, including data center locations.
- Incident date
- Dec 1, 2023
- Who
- Amazon
- Failure mode
- Hallucination
- AI surface
- Chatbot
- Severity
- Medium
What happened
Internal documents allegedly revealed that Amazon Q experienced severe hallucinations and leaked confidential data during its public preview. The leaked information reportedly included the locations of AWS data centers, internal discount programs, and unreleased features. Some employees categorized the incident as a sev 2, indicating a high level of urgency for engineering intervention.
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 system experienced large language model hallucinations, where the AI generated incorrect or fabricated information. These failures led to the unauthorized disclosure of sensitive internal data and the production of misleading answers regarding digital sovereignty.
What it cost
Sources
- PressAmazon’s Q has ‘severe hallucinations’ and leaks confidential data in public preview, employees warnplatformer.news
- PressAmazon's Q generative AI chatbot allegedly leaks location of AWS data centers - reportdatacenterdynamics.com
Cite this entry
https://failureindex.ai/failures/amazon-chatbot-allegedly-leaks-confidential-awsAI Failure Index. "Amazon Q chatbot allegedly leaks confidential AWS data and hallucinations" (FI-0401). Realm Labs. https://failureindex.ai/failures/amazon-chatbot-allegedly-leaks-confidential-aws (indexed Jun 10, 2026).Data fields CC-BY 4.0, prose citation permitted. Incident ID FI-0401. 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.