ChatGPT might have gone from friend to foe for Samsung

Samsung gave its employees access to ChatGPT a little over three weeks ago, intending to help them speed up various processes. Unfortunately for Samsung, sensitive information leaking three times already, as a few employees introduced this data into ChatGPT, forever becoming a part of the AI's learning database. And not just within the confines of Samsung's labs but everywhere someone uses ChatGPT. (via DigiTimes)

This sensitive information, including semiconductor equipment measurement data, among other things, became a part of the AI's network and language model outside of Samsung's grasp.

Here's everything that's known to have leaked

According to the report, the first leak happened when a Samsung employee at the Semiconductor and Device Solutions department came across an issue related to the semiconductor equipment measurement database. Looking for a quick solution, the employee entered the source code into ChatGPT, which means that Samsung's semiconductor equipment measurement database source code is now a part of ChatGPT's training data.

The second leak took place when another employee, trying to understand yields and other information better, entered code into ChatGPT and asked for it to be optimized.

And lastly, the third leak occurred when another employee instructed ChatGPT to create meeting minutes. Once again, all the information Samsung employees fed into ChatGPT became a part of the AI's knowledge base.

Upon hitting an average of one leak per week, Samsung took measures to prevent these sorts of things from happening again. The company reportedly instructed its employees to be careful about the data they share with ChatGPT. Samsung clarified that information fed to the AI chatbot gets transmitted to external servers where it can't be recovered or removed. The company also limited the capacity of each entry into ChatGPT to a maximum of 1,024 bytes.