Photo credit: iHeartMedia
iHeartMedia joins other companies like Spotify, Apple and Verizon in warning their employees against using ChatGPT.
An internal memo signed by iHeartMedia CEO Bob Pittman and CFO Rich Bressler set out expectations for OpenAI’s new technology. iHeartMedia informed employees that use of ChatGPT without prior approval should be severely restricted to prevent leakage of protected information.
The e-mail outlines a set of policies restricting the use of ChatGPT for iHeartMedia employees, including use on corporate devices or uploading corporate documents to such platforms. According to iHeartMedia, this step was taken to protect its intellectual property and other confidential information. That’s because ChatGPT stores conversations with users and uses them to train its AI algorithms. The large language model used in ChatGPT was trained by scraping data from the InternetWikipedia articles and other publicly available information online.
“While AI, including ChatGPT and other ‘conversational’ AIs, can be tremendously helpful and truly transformative, we want to be smart about implementing these tools to protect ourselves, our partners, our company’s information, and our user data,” reads the memo to iHeartMedia staff. “For example, if you upload iHeart information to an AI platform (like ChatGPT), that AI is effectively trained so that anyone – even our competitors – can use it, including all of our proprietary information.”
The memo outlines guidelines for using third-party AI tools, including consulting with iHeart’s legal and IT teams prior to implementing any use of the tool. “All projects require an assessment of the business impact and value of the project, a plan for monitoring and evaluation, and prior documented approval by legal and IT departments,” the memo said.
iHeartMedia is just one of many companies looking to protect their intellectual property from being inadvertently used to train LLMs like ChatGPT. Many of these companies have expressed concerns that these conversations could be used to leak confidential information. A practical example of this has already occurred Samsung employees have disclosed trade secrets via ChatGPT.