On August 17, 2026 (local time), OpenAI CEO Sam Altman mentioned a future ChatGPT memory feature in an interview with Z Fellows founder Cory Levy. The account Polymarket quoted it, and the phrase "perfect memory within 6 months" spread quickly on X.
What Exactly Did He Say?
Altman's original words were these: "In the next six months, we'll get close to a world where ChatGPT's next model sees your screen, records every meeting and call, and has perfect context on your entire life." He added, "I think we're about one model generation away from this actually becoming really useful."
You Do the Connecting
According to reports, Altman described this feature as a user-controlled structure that works only when users connect services themselves, such as texts, email, documents, and workplace collaboration tools. The point is that it doesn't mean the AI watches your screen without consent.
The Premise the Viral Clips Dropped
Short clips that spread on X compressed this remark into something like "within 6 months, AI will watch your entire life." Several outlets pointed out that the premise "users have to opt in to the connection" got lost in the process. The timing and content of the remark itself are real, but the framing that spread is stronger than the original.
Six Months Is a Forecast, Not a Promise
"Within six months" is Altman's personal outlook, not a confirmed release schedule. OpenAI hasn't published a separate official announcement or roadmap. It's also worth remembering that optimistic timing predictions from AI industry executives have often run ahead of actual releases.
One Setting to Check Now
If this feature does ship, the key question will be how finely users can control what gets connected. If you check now which items you've turned on and off in ChatGPT's memory and data connection settings, you can screen out unwanted connections in advance even when new features are added later.