Sam Altman Reveals GPT 6 Will Remember You
Fresh off the launch of GPT-5, OpenAI CEO Sam Altman is already looking toward the next horizon: GPT-6. In a recent meeting with reporters in San Francisco, Altman offered a rare look at the company's roadmap, suggesting the next iteration of its flagship AI will not only be different but will also arrive faster than its predecessor.
The Next Leap: GPT-6 and the Power of Memory
The defining feature of GPT-6, according to Altman, will be its ability to remember. He envisions an AI that moves beyond simple question-and-answer interactions to become a truly personal assistant. This means ChatGPT will need to learn and retain information about its users—their preferences, habits, and unique quirks—to provide a more tailored and adaptive experience.
"People want memory," Altman stated simply. "People want product features that require us to be able to understand them."
To navigate the complexities of creating such a personal AI, Altman revealed that OpenAI has been working closely with psychologists. This collaboration aims to shape the product thoughtfully, measuring how users feel and tracking their well-being over time. While this data hasn't been released, Altman hinted that it might be made public in the future.
Customization and Control: A More Adaptable AI
Altman also touched on the importance of ideological flexibility. He confirmed that future versions of ChatGPT will comply with a recent White House executive order requiring federal AI systems to be ideologically neutral yet customizable.
"I think our product should have a fairly center-of-the-road, middle stance, and then you should be able to push it pretty far," Altman said. He explained that if a user wants an AI that is "super woke," it should be able to provide that, just as it should be able to reflect a more conservative viewpoint if requested.
Learning from GPT-5's Rocky Rollout
Altman's forward-looking comments also came with a dose of reflection on the company's missteps. He acknowledged that the rollout of GPT-5 was "mishandled" after users took to social media to complain that the model felt colder and less helpful than GPT-4.
"I like the new one much better," he said, before noting that OpenAI had since pushed a quiet update to give GPT-5 a "much warmer" tone.
Navigating the Privacy Maze of a Personalized AI
While Altman called enhanced memory his favorite feature of the year, he was candid about the associated privacy challenges. A major concern is that temporary memory is not currently encrypted, which could leave sensitive user information vulnerable. He confirmed that adding encryption "very well could be" on the table, although a timeline has not been set.
He specifically pointed out that queries involving sensitive legal or medical information require robust protections that are not yet in place. "If you can get better versions of those from AI," he said, "you ought to be able to have the same protection for the same reason we decided you could get them from a doctor or a lawyer."
Beyond Chat: Altman's Vision for the Future
Looking even further ahead, Altman shared his interest in technologies adjacent to AI, calling "neural interfaces a cool idea." He imagines a future where a user could simply "think something and have ChatGPT respond."
While that future may be a way off, Altman's current focus remains on making OpenAI's core product more useful in daily life, from work tasks to parenting questions. However, he ended with a surprising note of caution, suggesting that the current model of interaction has its limits.
"The models have already saturated the chat use case," Altman said. "They're not going to get much better. ... And maybe they're going to get worse." This suggests that a true evolution, like the one promised with GPT-6's memory, is not just a feature—it's a necessity.
