ChatGPT has a “memory” now, critics say it’s just OpenAI storing your data

ChatGPT will now be able to memorize your previous chats.

On February 13th, OpenAI started adding its new “Memory” feature to ChatGPT. With the new feature, the chatbot will be able to remember what users have previously said and use that information in future chats.

The feature is currently in a test phase. OpenAI’s custom chatbots, named GPTs, will also have the memory feature. “You’ve told ChatGPT you own a neighborhood coffee shop. When brainstorming messaging for a social post celebrating a new location, ChatGPT knows where to start,” explains OpenAI on how the new feature works.

Users can configure ChatGPT’s settings to customize how the chatbot remembers their conversations or can turn it off completely. According to the company, users are in control of making the chatbot forget something.

Privacy concerns

ChatGPT learning and understanding all about its users’ history and habits is causing privacy concerns. While users can opt out, the memory feature is switched on by default. There have already been concerns about how the company is using the data that’s fed to the chatbot.

OpenAI has already been under fire for privacy issues. In 2023, Italy temporarily banned ChatGPT for the use of personal data to train its models without user consent. This is in violation of the EU’s data protection and privacy laws known as GDPR.

In a blog post, OpenAI agrees that the memory feature brings additional privacy and safety considerations.

“We’re taking steps to assess and mitigate biases and steer ChatGPT away from proactively remembering sensitive information, like your health details – unless you explicitly ask it to,” the company explained.

OpenAI also added a ‘Temporary chat’ option to ChatGPT, allowing users to have quick conversations that aren’t memorized.

Don’t call it a memory

While giving an AI such anthropomorphic features as a memory sounds intriguing, critics warn against labeling it as such. Paris Marx, a Canadian technology journalist, wrote on X that the media should stop repeating tech companies’ misleading framings of their products.

“ChatGPT does not have a “memory.” It does not “remember” anything. OpenAI is just storing data about you and your engagement with its product,” he said.

“For those who don’t see the issue: the anthropomorphism misleads people about what the tech does. It plays into the misconception that the chatbot is on the verge of replicating human intelligence. If you’ve been deceived about what the tech is, you can’t respond to it properly,” he adds.

AI researcher Joscha Bach opposed Marx, saying that phrases like memory have been used for a long time in the technical context. For example, a computer’s storage capabilities are described as “computer memory.”

“ChatGPT stores data that allows it to recreate the content of past interactions. That’s pretty much memory,” says Bach.