ChatGPT has become a headache for Samsung. The popularity of the AI-powered chatbot has prompted some companies to start integrating it into their workflows. One was Samsung Semiconductor, where engineers used it to fix bugs in their product code, unaware they were passing sensitive information.
According to the report Economist , Samsung employees leaked confidential content three times, some of which included source code. The incidents occurred less than twenty days after the South Korean giant authorized the use of ChatGPT in its semiconductor division. According to the publication, employees sought to take advantage of two key features of the chatbot, unaware that the data would fall into OpenAI.
The first leak came from an engineer in charge of their own database software. A member of staff encountered a compilation error and simply copied and pasted the source code into ChatGPT to find a solution. While many turn to chatbots to solve scheduling problems, OpenAI encourages users not to enter sensitive information.
The second leak came when another employee entered a code used to identify faulty chips and asked ChatGPT to optimize it. This leak can be considered the most sensitive due to the secrecy of Samsung’s manufacturing processes. Finally, a third employee entered the meeting details document for the AI to generate the presentation.
Faced with the leaks, Samsung took emergency measures and imposed a download limit of 1024 bytes on ChatGPT issues. The company confirmed that it will investigate and punish those responsible if necessary.
ChatGPT is a headache for some companies
Samsung’s confidential information leak may not be the first ChatGPT has received. Since its explosion in popularity, thousands of users have turned to OpenAI’s artificial intelligence to ask questions, solve problems, and improve content. Some programmers grab source code while waiting for the AI to find bugs or optimize the program.
Few people know that all the information they enter into ChatGPT ends up on OpenAI servers and is used to train the model. This has already raised alarm bells with many companies restricting or prohibiting their employees from using the chatbot. Amazon has already issued warnings to its employees, while Verizon and JP Morgan Chase have chosen to ban use over fears of a leak.
Samsung’s case is significant because the sensitive data came from one of its largest divisions. Samsung Semiconductor is responsible for manufacturing the processors, image sensors, memory, and storage solutions used in both its products and other third-party devices. The bad chip detection model requires special shielding in a competitive industry like the semiconductor industry.
According to the report, Samsung will wait for the end of the investigation to make a decision. The company allows some of its employees to use ChatGPT to introduce them to the technology, although it may ban it if the leaks continue. Departments such as mobile communications, user services, and smart home departments are not allowed to use it in their workgroups.
See also: Bill Gates says it’s a bad idea to contain more powerful AI models than GPT-4.
