Open AI says the data is used to train AI models and improve their responses, but its terms allow the company to share users’ personal information with affiliates, vendors, service providers and law enforcement. “So it’s hard to know where your data is going to end up,” Love said.
OpenAI’s privacy policy says ChatGPT collects information to create accounts and communicate with companies, said Bharath Thota, data scientist and chief solutions officer for analytics at Kearney, a management consulting firm that advises companies on managing and using AI data to drive new revenue streams.
This data collection may include name, account credentials, payment card details and transaction history, he said. “Personal information may also be stored, particularly if an image is uploaded as part of the prompt. Similarly, if a user decides to connect with a company’s social media pages such as Facebook, LinkedIn or Instagram, personal information may be collected if they share their contact details.”
OpenAI uses consumer data like other big tech and social media companies, but it doesn’t sell ads; instead it offers tools. That’s an important distinction, says Jeff Schwartzentruber, senior machine learning scientist at security firm eSentire. “User-input data is not used directly as a commodity. Instead, it’s used to improve services that benefit users, but also increase the value of OpenAI’s intellectual property.”
Privacy Controls
Amid criticism and privacy scandals since its founding in 2020, OpenAI has introduced tools and controls that can be used to lock down data, saying it is “committed to protecting people’s privacy.”
Regarding ChatGPT, OpenAI said it understands that users may not want their information to be used to improve models, which is why it offers ways for them to manage their data: “ChatGPT Free and Plus users can easily control whether they contribute to future model improvements in their settings,” the company wrote on its website, adding that it doesn’t train with API, ChatGPT Enterprise, or ChatGPT Team customer data by default.
“We offer ChatGPT users several privacy controls, including an easy way to opt out of training our AI models and an ephemeral chat mode that automatically deletes chats at regular intervals,” OpenAI spokesperson Taya Christianson told WIRED.
The company says it doesn’t ask for personal information to train its models, doesn’t use publicly available information on the internet to create profiles of people, doesn’t advertise to them, doesn’t target them, and doesn’t sell user data.
OpenAI doesn’t train its models on audio clips from voice chats, unless you choose to share the audio “to improve voice chat for everyone,” according to a voice chat FAQ on the OpenAI website.
“If you share your audio with us, we may use audio from your voice chats to train our models,” Open AI said in its FAQ on voice chats. Meanwhile, transcribed chats may be used to train models, depending on your choices and plans.
