OpenAI is giving ChatGPT something it has long lacked: a lasting memory of its users. The move promises more personalized help over time, while sharpening debates about data, control, and regulation.

Early research and “dreaming” (June 3)

On June 3, OpenAI detailed internal work on a system it calls “Dreaming,” aimed at “better memory for a more helpful ChatGPT,” focused on “freshness, continuity and relevance.”¹ This research underpins how ChatGPT now synthesizes information from prior chats into usable long‑term memory.

Rollout of the upgraded memory system (June 4)

By June 4, the company began deploying those advances in production. Tech outlet The Verge reported that “ChatGPT's upgraded memory system is rolling out to everyone,” describing it as an enhancement of the earlier background “dreaming” feature that lets ChatGPT “sort through your conversations and save information in the background.”² According to the report, ChatGPT Plus and Pro subscribers can access the upgrade immediately, with free users to follow in the coming weeks.²

From feature to life‑long assistant (June 5)

A second Verge article explained how the new long‑term memory works in practice: ChatGPT can “reference details you discussed, even if you didn’t manually save them” and will now recall information via both explicit “saved memories” and “reference chat history,” the latter being “insights ChatGPT gathers from past chats to improve future ones,” according to OpenAI.³ CEO Sam Altman said the chatbot can “now reference all your past conversations,” aligning with a goal to build “AI systems that get to know you over your life.”³

The feature is optional: users can disable it in personalization settings or use temporary chats that do not affect memory.³ The upgrade is initially limited to paid users and excludes the EU, UK, Switzerland, Norway, Iceland, and Liechtenstein, regions noted for stricter AI rules that Altman has previously criticized.³

Human and AI perspectives

From OpenAI’s side, leadership framed the change as a major capability jump. Altman announced a “big upgrade to chatgpt memory rolling out today!” on X,Sam Altman (sama)⁴ amplifying OpenAI’s own message that it has been “researching new ways for ChatGPT memory to carry context across conversations and keep it useful over time,” now shipping as “a more capable memory system in ChatGPT.”Sam Altman (sama)⁵

Researchers characterize the same shift more technically, emphasizing that the improved memory synthesis should make ChatGPT more consistently helpful by maintaining “continuity” across sessions while keeping information up‑to‑date and relevant.¹ Human commentators, meanwhile, highlight both convenience—an assistant that remembers preferences and past tasks—and the unresolved questions of how such long‑term data is stored, governed, and regulated over a user’s digital lifetime.