Credit: VentureBeat created with ChatGPT
Openai slowly rolls out a better memory for ChatGPT. This makes it default for ChatGPT reference past conversations. This has caused users to be uncomfortable about how much information the platform knows.
ChatGPT logs information about previous interactions via its Memory feature. This ensures that preferences are saved, and conversations can continue seamlessly from where the user left them off.
With this new update, ChatGPT can “draw on previous conversations to deliver more useful and relevant responses” across all platforms. Memory improvements allow future conversations to refer back to previous chats, not just the current chat window. This feature will only be available to ChatGPT Plus users and Pro users. ChatGPT Enterprise Team and Edu users will have access to this feature later.
Memory in ChatGPT now allows you to reference all your previous chats for more personalized responses. It will also take into account your preferences and interests, making it more useful for writing, learning, getting advice and more. pic.twitter.com/s9BrWl94iY
– OPENAI (@OpenAI)””https://twitter.com/OpenAI/status/1910378768172212636?ref_src=twsrc%5Etfw””> OpenAI added Memory in February of last year to ChatGPT to make it easier to talk to ChatGPT. Memory is a feature that most chat platforms, as well as large language models (LLMs), offer. Gemini 2.0 Flash Thinking has added Memory. Frameworks like A-Mem enhance long-context memories for more complex tasks.
OpenAI says that improvements to memory will allow ChatGPT to “naturally build on” earlier chats, and over time, interactions on ChatGPT should be more tailored to each user. OpenAI offers two options to control Memory via settings. Reference Saved Memories is the first option, where users can instruct ChatGPT to memorize facts such as names or preferences. The company stated that people add this information when they explicitly tell ChatGPT to keep something in mind. The model will determine which information will be useful in future conversations.
Reference Chat History is the second control. This setting allows ChatGPT “adapt” to your tone, goals and interests by drawing context from previous discussions. However, the context is not stored or displayed in the settings page, like saved memories. OpenAI stated that you can turn both settings on or off or only reference saved memories. “The settings can be changed at any time, including managing specific saved memorys. ChatGPT will not use past conversations if you opt out. You can also ask it what it remembers, or switch to Temporary Chat to memory-free sessions.”
Concerns of some users
Having access to preferences and context for enterprise tasks makes AI models more valuable.
AI Investor Allie K. Miller stated in A post on X said that this update made ChatGPT “listen all the time.” It is a feature that cuts across all your conversations, regardless of whether you explicitly asked it to do so or not.
As I said a few weeks back, memory is arguably the best feature within these platforms. As models and features become commoditized it will come down to personalization and collaboration. Memory is key. Miller said, “Memory is the moat.”
I’ll explain the brand new OpenAI release that kept Sam Altman awake all night yesterday.
I haven’t turned memory on ChatGPT because I can’t do it for work, and today’s the day I start to feel it.
ChatGPT’s memories were pretty boring until today. It waited until a clear… https://t.co/fdAq1JLlwA
– Allie K. Miller (@alliekmiller)””https://twitter.com/alliekmiller/status/1910385168415940881?ref_src=twsrc%5Etfw””> April 10, 2025
Prominent AI expert and Wharton Professor Ethan Mollick Noted that it’s not something he will activate.
I totally understand why AI long-term memories are useful and, based upon my testing, I think many people will enjoy it… But I don’t actually want my LLMs that I use for my work to add personal details or subtly alter its answers as a consequence of my previous interactions. Mollick said that boundaries are important.
Although I understand why AI long-term memory is useful, and based on my tests, I think many people will enjoy it… I don’t actually want my LLMs that I use for my work to add personal details or subtly alter its answers based on past interactions.
Boundaries can be good.
Ethan Mollick @emollick April 10, 2025
OpenAI Cofounder Andrej Karpathy””https://x.com/karpathy/status/1910411355300954539″ ” rel=””noreferrer noopener”” target=””_blank” “> ChatGPT “think less of me because you asked that noob bashing question 7 months ago.”
Will GPT think less of me because I asked that noob bashing question 7 months ago?
– Andrej Karpathy (@karpathy)””https://twitter.com/karpathy/status/1910411355300954539?ref_src=twsrc%5Etfw””> April 10, 2025
The ChatGPT memory will be useful, but the user will have to decide how much information they want to share with the chat platform and how important past conversations will be.
VB Daily provides daily insights on business use-casesWant to impress your boss? VB Daily can help. We provide you with the inside scoop on what companies do with generative AI. From regulatory shifts to practical implementations, we give you the insights you need to maximize ROI.
Read our privacy policy
Thank you for subscribing. Click here to view more VB Newsletters.
An error occured.
