OpenAI tests ChatGPT “memory,” opening up a can of privacy worms — and possibly a new revenue source?

The bot that knows you best… The world’s most famous chatbot is going to remember you. For those tired of repeating the same info to ChatGPT, OpenAI introduced a solution: a memory feature that’ll store info about you to personalize responses. Like: it’ll recall that you’re lactose-intolerant and stop sharing cheese-loaded recipes. The feature, which is being tested with a small number of users, will be turned on by default (though folks can opt out).

• No reminder needed: The goal is for CGPT to get smarter and more personalized. For example, the feature will allow users of Books GPT (a custom bot on the GPT store) to get reading suggestions based on past conversations and genres.

• Privacy flags raised: OpenAI said it has trained CGPT not to remember sensitive info like health-related convos. Still, the feature may stoke concerns — especially because OpenAI said memory data would be used to train its models.

OpenAI-ing another can of worms… While memory should make CGPT more convenient to use, it raises privacy concerns at a time when US and EU regulators are cracking down on tech giants’ data-tracking practices. It could also heighten (already loud) alarm bells among consumers, regulators, and lawmakers (who are drafting AI legislation). In the US, 400+ AI bills have been introduced, while the EU is set to pass a historic AI Act. Meantime, copyright-infringement suits are stacking up.

• New player, same moves? OpenAI’s memory feature resembles the tracking that companies like Meta, Google, and Amazon have been doing for years. They learn how you use their services and build digital profiles… often for ad targeting.