Reversal
Samsung and the code that went to ChatGPT
Samsung Electronics · Electronics · South Korea
Weeks after letting engineers use ChatGPT, Samsung banned generative AI on company devices — staff had pasted confidential source code into it. An all-in that opened and closed in about three weeks.
Allowed
generative AI opened to staff
Banned
after source-code leaks to ChatGPT
The record
- In late April 2023, Samsung Electronics told staff in its device-solutions division that generative-AI tools (ChatGPT, Bard, Bing) were banned on company-owned devices, effective 1 May 2023. (Bloomberg, 2023-05-02; Reuters)
- The ban followed multiple instances of employees uploading sensitive internal source code and materials to ChatGPT. (Bloomberg; Forbes)
- Samsung warned staff that data entered into such tools is stored on external servers and is difficult to retrieve or delete. (Bloomberg)
Three weeks
The division had opened generative-AI tools to engineers. Within about three weeks, sensitive internal source code had been pasted into ChatGPT in more than one instance, and the tools were banned on company devices effective 1 May 2023. Open door to closed door, in under a month.
It reads as a story about a leak. It is really a story about a boundary that was never drawn before the door was opened.
The escape hatch that wasn’t there
The failure wasn’t the model; it was deploying a third-party tool into a confidential workflow with no data-governance boundary around it. Once code is pasted into an external service, it is on someone else’s servers, and it cannot be un-sent. Enthusiasm and “try it” had quietly become the same decision as “we accept this data leaving.”
Samsung’s cost was a near-miss — no fine, no quantified loss reported — which is exactly why it is instructive. The correction came before the damage compounded, because someone noticed the boundary was missing while the stakes were still recoverable.
The lesson
Enthusiasm for a tool is not a deployment plan. Without a data-governance boundary, “let’s try it” and “we can’t un-send that” are the same decision.
How we’re reading this
This is a near-miss, not a costed failure — no fine or quantified loss was reported. The precise trigger (a single engineer or several) is contested; we say “multiple leak instances.” Samsung did not abandon AI and later built internal tools; we report the ban and its stated cause.
Sources
- 01Samsung Bans Staff’s AI Use After Spotting ChatGPT Data Leak
Bloomberg · May 2, 2023
- 02Samsung Bans ChatGPT Among Employees After Sensitive Code Leak
Forbes · May 2, 2023
Compiled from public filings, court records, company statements and reputable press. Figures are attributed to their sources; allegations are labeled as such. Not legal or investment advice.