The recently introduced Amazon Q chatbot is encountering significant issues such as "hallucinations" and leaks of confidential data, as reported by Platformer. Users of the preview version gained access to information about the locations of the company's data centers (DCs), undisclosed discount programs, and unreleased features.
The incident has been classified as a level 2 leak, indicating its severity and prompting specialists to address the problem during nighttime hours and weekends. Amazon Q's problems come at a critical time as Amazon aims to secure its position in the AI assistant market. Microsoft, Google, and several startups have been pioneers in this field. In September, the company announced plans to invest $4 billion in the startup Anthropic, and this week, it finally unveiled its own chatbot, Q.
In response to reports of issues with Amazon Q, the company stated that some employees do share feedback through internal channels, but there is no evidence that feedback materials mention instances of explicit data leakage. Later, the company reiterated that Amazon Q did not permit leaks of confidential information. However, internal documents leaked online suggest otherwise: Amazon Q provided misleading or potentially hazardous responses, particularly concerning security issues.
Until now, Amazon has diligently concealed the locations of its DCs within its extensive portfolio, encompassing both owned facilities and leased capacities. In 2018, WikiLeaks revealed internal documentation of the IT giant, including coordinates of the company's data centers as of 2015.
Since then, DC locations have only been disclosed in documents related to planned construction, with precise information about the whereabouts of all company facilities still unavailable. In the latest K-10 report, AWS reported that by the end of 2022, the company owned DCs worldwide with a total area exceeding 1.4 million square meters and leased an additional 1.6 million square meters.