OpenAI rolled out a fresh update to the ChatGPT app this week, and if you thought the bot already knew too much about you, buckle up.
The company says the new version lets the assistant pull from your entire chat history to personalize responses, which is corporate-speak for "it remembers that embarrassing thing you asked at 2 a.m. in March." The feature is rolling out to Plus and Pro subscribers first, because of course it is.
Free users get to watch from the cheap seats, as is tradition in the great Silicon Valley paywall industrial complex.
Here's the gist: instead of starting every conversation from scratch like a stranger with amnesia, ChatGPT can now reference past chats, your stated preferences, and whatever random facts you've fed it over the months.
In theory, this means fewer repetitions and more tailored answers.
In practice, it means your AI buddy is quietly building a dossier that would make your therapist raise an eyebrow.
OpenAI frames this as a convenience play.
Ask it for dinner ideas and it'll remember you hate mushrooms.
Tell it you're training for a marathon and it'll stop suggesting couch-based hobbies.
The problem is that "memory" in AI is a loaded word, and the company's track record on explaining exactly what gets stored and where is, let's say, a work in progress.
Privacy folks are already sharpening their keyboards.
The EU has been poking at OpenAI for months, and any feature that retains more user data is basically catnip for regulators.
On this side of the pond, the FTC has shown it's not shy about asking pointed questions when tech companies get fuzzy about data practices.
For the average user, the practical question is simpler: does this make the app better or just nosier?
Early testers report the memory feature is genuinely useful for ongoing projects, like drafting a novel or managing a budget spreadsheet.
It keeps context, which is the whole point of paying twenty bucks a month.
The flip side is that clearing your history now feels less like tidying up and more like shredding evidence.
You can turn memory off in settings, which OpenAI has been careful to advertise loudly.
Whether that toggle actually wipes everything or just hides it from the chat window is the kind of detail that deserves more than a help-center paragraph.
Transparency has never been the company's strong suit, and a feature this personal demands better than vague assurances.
Google's Gemini and Anthropic's Claude have their own memory experiments in the works, and Apple keeps whispering about on-device processing as its privacy calling card.
The arms race isn't really about who's smartest anymore.
It's about who can remember the most without creeping everyone out, which is a weird finish line for an industry that used to sell us on "don't be evil." If you're a paying subscriber, the update will show up whether you asked for it or not.
Dig through the settings, decide what you're comfortable with, and maybe think twice before confessing your weirdest shower thoughts to a server farm in Iowa.
The memory upgrade is genuinely handy, but "handy" and "unsettling" aren't mutually exclusive.
OpenAI built a tool that listens, remembers, and never forgets unless you tell it to, then acted surprised when people felt weird about it.
Final Thoughts
That's just a group chat with better grammar.