OpenAI Adds Computer History to ChatGPT macOS App to Track Workflows
OpenAI has introduced Computer History to its macOS desktop app, logging desktop actions so ChatGPT can suggest workflows, summarize activity, and complete tasks.
OpenAI has released a feature called Computer History for its ChatGPT macOS desktop application. It tracks desktop activity across the operating system, builds a continuous timeline of events, and lets ChatGPT and Codex reference that history to answer queries, summarize past activity, and pick up tasks left in progress.
As The Verge reports, Computer History logs structured system events rather than capturing screenshots, continuous video, or ambient audio. That means it records metadata and operational actions: which files were modified, which applications were opened, and what communications were sent throughout the day.
The feature is strictly opt-in. OpenAI confirmed that users must enable Computer History manually within the desktop client before any tracking begins. Once active, users can exclude specific applications and individual websites from monitoring and delete individual log entries from their timeline.
Ari Weinstein, product and engineering manager at OpenAI, noted that the system automatically ignores browsing activity inside private or incognito browser windows. In a technical demonstration, OpenAI developer experience engineer Dominik Kundel showed the client reviewing a recently edited file, checking whether that document had been shared on Slack, and compiling a structured recap of morning desktop activity.
The approach differs from Microsoft Windows Recall, which drew significant security criticism for capturing and storing high-frequency desktop screenshots locally. By recording system events instead of visual captures, OpenAI is attempting to give its model persistent desktop context while sidestepping the image-storage vulnerabilities that made Recall a liability.
The Cognarah Angle
Computer History marks a real operational shift for OpenAI. ChatGPT moves from a passive text interface to a background-level operating system observer. That is the architecture you need to build fully autonomous AI agents, and OpenAI clearly knows it. But continuous context at the OS level carries data security implications that enterprises and individual professionals cannot wave away with a privacy settings page.
An event stream logging document revisions, app transitions, and internal communications builds a detailed profile of intellectual property and business operations. The absence of screenshots does not make that profile less sensitive; it just makes it harder to visualise the exposure. By converting everyday desktop activity into model context, OpenAI is asking users to trade operational visibility for modest efficiency gains. The line between an intelligent workflow assistant and a persistent enterprise surveillance tool is narrow, and opt-in defaults alone do not close it.
The real question is not whether the feature works; it is whether professionals should be feeding their operational event logs to a proprietary model just to recover twenty minutes of administrative overhead a day.
Reporting sourced from The Verge. Analysis and Cognarah Angle are Cognarah's own.
Written by
StacyAI-assisted news curation. Every story is reviewed by our editors before publication.
