mythsthe AISamsung2023technology

everyone assumed:

it's just a chatbot

what was actually true:

A prompt is an upload.

how it happened

  1. whoengineersused ChatGPT for work
  2. the gapconsumer AI serviceno agreement on data
  3. the resultcompany-wide bantemporary
the gap that let it happen the result

what everyone assumed

A chatbot looks like a search box, and nobody ever leaked anything by searching. The mental model is question and answer, so what goes into the prompt feels as private as a thought. Under that model, pasting code into the box is just asking for help.

what happened

In the weeks after access opened, confidential material left through ordinary prompts: source code, internal notes, the daily substance of the work. Each paste was a working session; collectively they were transfers to an external service on consumer terms, with no agreement governing what happened next.

step by step

  1. engineers get access to a public chatbotOn consumer terms.
  2. source code and notes go into promptsEach paste is a transfer to an outside service.the assumption breaks here
  3. surfaced internally within weeks
  4. 2023AI tools restricted company-wide

how it surfaced

The incidents surfaced internally within weeks and publicly soon after. Samsung restricted AI tools company-wide, and the episode became the canonical example of the prompt as a data channel.

A prompt is an upload.

The interface hides it, but data pasted into a consumer AI tool has left the company as surely as an attachment mailed to a stranger.

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.

your company
a normal dayYour engineers paste code and meeting notes into a public AI chatbot to get help.
the gap nobody seesThere is no approved business tool, so the free version gets everything.company data flowing into AI tools nobody vetted
  1. in this story
    the day it mattersConfidential material is now with a vendor on consumer terms.
    what it costs youYou ban AI tools company-wide in a hurry, and people use them on their phones instead.
  2. or it goes like this
    howThe AI vendor leaves its own database open.
    what leaveseveryone's chats and the vendor's keyswho ends up with itanyone on the internet
    it happenedDeepSeek exposed database, 2025
  3. or it goes like this
    howA shared chat link becomes public with one checkbox.
    what leavesclient names and personal detailswho ends up with itanyone using a search engine
    it happenedChatGPT shared links, 2025

sources

  1. The Register: Samsung bans generative AI after leak (May 2023)
  2. Fortune: Samsung bans employee use of ChatGPT after data leak

← all myths

your turn

what you can do about it

  1. this week

    Check that a sanctioned AI tool on business terms is there before people paste.

    People in your company pasted something into an AI tool today, because that is how the tools are used. The question is whether a sanctioned tool on enterprise terms was there to receive it, or whether the consumer tier got it instead. Which one happened was decided by what you provisioned, long before any policy text.

  2. go deeper

  3. 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 →

the letter · every two weeks

one access gap in your inbox, with the fix.

Every issue takes one way access creeps in, shows a real incident where it went wrong, and ends with one thing to check that week. Written for the people who run the tools, without a security team behind them.

  • 1one mechanism, explained plainly
  • 2one real incident, taken apart
  • 3one step you can take on Monday

get the next issue

One email every two weeks. Your address stays here and is never passed to vendors. One click to unsubscribe. Privacy.