Less than three weeks after Samsung Electronics lifted its internal ban on ChatGPT on March 11, 2023 to boost productivity, employees in the chaebol's device solution and semiconductor business unit reportedly leaked confidential corporate data into the service at least three times, according to a Korean media report relayed by The Register on April 6, 2023. In the first incident, an engineer pasted the entire source code of a problematic semiconductor-database download program into ChatGPT and asked the model for a fix; a second employee uploaded program code designed to identify defective semiconductor equipment; a third uploaded a recorded internal meeting in an attempt to auto-generate minutes. Because ChatGPT's FAQ at the time warned that conversations may be reviewed by OpenAI trainers to improve its systems, the inputs were potentially exposed to OpenAI staff and to the model's training data. Samsung responded with "emergency measures" that capped uploads to 1,024 bytes per prompt and told employees that any repeat incident could trigger a full network-level block of ChatGPT, and local reports indicated Samsung was considering building its own in-house AI service. The episode also rattled domestic peers: SK hynix blocked chatbot use on its internal network pending security approval, and LG Display launched an employee education campaign.
Blog
Insights on AI security and development
Learn how to protect your code, understand AI risks, and build secure workflows with expert insights.

