how it happened
- whoanyone onlineno credentials needed
- the gapexposed log storequeryable from a browser
- the result1M+ log lineschat history and secret keys
what everyone assumed
Sending a prompt to an established-looking AI tool feels like handing it to a competent operator who will protect it. The assumption is that the vendor has secured the back end, because securing the back end is the vendor’s basic job.
what happened
The exposed database accepted queries from anyone, returning plaintext chat history, API keys, and backend secrets through a browser. The prompts users assumed were protected were sitting in an open store, alongside the keys to the service itself.
step by step
- a database is left open to the internetIt accepts queries from anyone, no credentials needed.the assumption breaks here
- chat history and secrets sit insidePlaintext prompts, API keys and backend secrets.
- January 2025Wiz Research finds and discloses itOver a million log lines reachable through a browser.
how it surfaced
In January 2025 Wiz Research disclosed the open DeepSeek database: over a million log lines including plaintext chat history and secret keys, reachable with no credentials.
The security of your prompts is exactly as good as the vendor’s weakest configuration, which you cannot see and did not vet.
Trusting a tool with data is trusting an operation you have not inspected.
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 mattersResearchers find the vendor's database open to the internet, chats included.what it costs youYour client notes were readable by anyone who looked, and you never chose that vendor.
- 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 happenedChatGPT shared links, 2025howA shared chat link becomes public with one checkbox.what leavesclient names and personal detailswho ends up with itanyone using a search engine
sources
your turn
what you can do about it
this week
List the AI tools that hold your company's data, and whether anyone vetted the vendor.
Your teams send company information into AI tools daily, on the assumption the vendor has locked the back end. DeepSeek is proof that assumption can be entirely wrong. Which AI tools hold your data, and whether anyone checked the vendor before the data started flowing, is the gap here.
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 →