OpenAI has introduced a "Computer History" feature for the ChatGPT desktop application on macOS that records user clicks and keystrokes [1, 3].
The tool allows the AI to monitor a user's digital behavior in real time. By tracking app switches and web browsing, the system builds a searchable timeline that the AI can reference to provide more personalized assistance [1, 3, 4].
OpenAI designed the feature to help the AI understand how a user works [1, 2]. This data allows the assistant to suggest specific automations or pick up tasks that a user left unfinished [1, 2]. The goal is to move the AI from a reactive chat interface to a proactive assistant that understands the context of a user's entire workstation [1].
Privacy concerns have emerged regarding how this sensitive data is handled. Digital Trends said the feature is a more private, screenshot-free upgrade compared to similar tools like Chronicle [2]. However, The Next Web said that Computer History stores the resulting memory files without encryption [3].
The rollout for macOS users began earlier this year [1, 5]. The feature integrates directly into the desktop experience, allowing the AI to bridge the gap between different software applications by knowing which tools the user is currently employing [1, 3].
Because the system tracks granular inputs, including keystrokes, it creates a comprehensive log of user activity. This timeline serves as a long-term memory for the AI, ensuring it does not lose context when a user switches between a web browser, a document editor, or a coding environment [1, 4].
“The system builds a searchable timeline that the AI can reference to provide more personalized assistance.”
This shift represents a transition toward 'agentic' AI, where the software moves from answering prompts to observing and acting upon a user's entire operating system. While this increases productivity through automation, the lack of encryption for local memory files creates a significant security vulnerability if a device is compromised, as a complete log of a user's digital life would be accessible in plain text.
