how it happened
- whoenterprise launch"more secure than consumer AI"
- the gapAI assistanthallucinations, confidential details
- the resultleaked ticketsAmazon disputes any leak
what everyone assumed
When a tool arrives wearing the word enterprise, the natural reading is that it has been built to a higher safety bar than the consumer version of the same idea. That is a reasonable inference, because enterprise usually does mean more controls, more guarantees, more review. The label invites you to relax the caution you had been applying to the consumer chatbot, on the assumption that the careful version has already done the worrying for you.
what happened
Here the quiet stretch was only days, the window between a confident launch and the first internal warnings. According to leaked documents reported by Platformer, employees flagged that the assistant was producing severe hallucinations, and warned that it could surface confidential material such as the location of data centres and unreleased features. One of the incidents was marked at a severity that meant engineers were paged overnight. Amazon disputes the central claim. A spokesperson said the assistant had not leaked confidential information, described the reports as employees sharing feedback through normal internal channels, and said no security issue was identified as a result.
step by step
- Amazon Q launches as the more secure optionMarketed to enterprises as a safer assistant.
- employees flag severe hallucinationsLeaked tickets warn it could surface confidential material. One incident pages engineers overnight.the assumption breaks here
- three days after launchthe tickets are reported publiclyAmazon disputes the claim and says no security issue was identified.
how it surfaced
The warnings became public when the internal tickets were reported, three days after the launch that had called the tool more secure than the alternative. What retired the assumption was not a confirmed breach, it was the gap itself: the enterprise label and the employees raising their hands at the same moment, with the company saying one thing and the leaked tickets saying another. The label had promised a settled question. The question was not settled.
Enterprise describes who a tool is sold to, not how it behaves under load.
The word is a market segment, and it does not, on its own, tell you what the model will say or what it can reach.
picture it at a company like yours
The company here is made up. The other two paths are not: they happened, and each links to the story.
- in this storythe day it mattersStaff notice it confidently invents answers and repeats details from documents it should not show them.what it costs youThe tool you bought to be careful is now the one leaking and guessing, and the contract says it is secure.
- or it goes like thisit happenedSamsung, 2023howEngineers paste code and notes into a free chatbot to get help.what leavesconfidential code and meeting noteswho ends up with itan AI vendor on consumer terms
- or it goes like thisit happenedDeepSeek exposed database, 2025howThe AI vendor leaves its own database open.what leaveseveryone's chats and the vendor's keyswho ends up with itanyone on the internet
sources
your turn
what you can do about it
this week
Take one enterprise AI tool your team relies on and find the actual control behind the label: what data it can see, what it is allowed to repeat, and who would hear about it first if it got something wrong.
Somewhere in your company a tool is trusted a little more than its consumer cousin because of the tier it was bought on. The question the quiet days ask is whether that extra trust was earned by something you can point to.
go deeper
if you're not sure where to start
Run the offboarding checklist on one leaver. It covers every account and token to close when someone leaves.
open the checklist →