OpenAI Launches Computer History for macOS to Track User Workflows

OpenAI has released "Computer History," a new opt-in feature for the ChatGPT Mac app designed to give the AI a comprehensive understanding of a user's digital activity. Available to Pro, Business, and Enterprise users, the tool creates a searchable timeline of interactions that ChatGPT can reference to provide context-aware assistance.

Moving away from the screenshot-based approach used in the earlier "Chronicle" research preview, Computer History utilizes macOS accessibility systems to record specific interaction events. According to OpenAI, the system tracks clicks, typing, keyboard shortcuts, and application switches across selected apps and websites. To function, the feature must be manually enabled by the user and requires the "Memories" setting to be active.

A Shift in Data Collection

In a bid to address privacy concerns associated with constant screen recording, OpenAI confirmed that Computer History does not capture screenshots, screen recordings, microphone input, or system audio. Instead, it generates an interaction-event stream based on data exposed through the macOS accessibility system.

Data handling for the feature involves a hybrid local-cloud model. Interaction events are stored on the local machine for up to 48 hours. These events are then processed on OpenAI's servers to generate "memories," though the company claims these events are not retained after processing nor used for model training. The resulting memory files are stored locally as plain-text Markdown files located at ~/.codex/memories/extensions/skysight/.

The Push for OS-Level AI

This launch signals OpenAI's broader strategy to integrate ChatGPT deeper into the operating system, transforming the AI from a standalone chatbot into a personal assistant with full workflow context. By tracking how a user moves between documents and apps, the AI can answer complex queries about past activity, such as identifying exactly what a user was working on prior to a break.

However, this shift toward OS-level awareness mirrors controversial industry trends, such as Microsoft's Recall. While the productivity gains in document retrieval and workflow automation are significant, the implementation introduces new security considerations. Specifically, the storage of activity logs in unencrypted local files creates a potential vector where other programs running under the same macOS user account could access sensitive history.

What to Watch

As OpenAI continues to refine its integration with macOS, the industry will be watching how users balance the utility of a total-recall assistant against the security risks of local activity logging. It remains to be seen if OpenAI will introduce encryption for the local Markdown memory files or expand the feature to other operating systems to achieve a cross-platform workflow timeline.