OpenAI Launches Computer History: A Local, Searchable Timeline of macOS Activity for ChatGPT Memory

OpenAI has released Computer History, a macOS feature that converts user activity—clicks, keystrokes, keyboard shortcuts, and app switches—into a searchable memory timeline that ChatGPT and Codex can use as context. The feature replaces an earlier research preview called "Chronicle," which relied on screenshots.

Computer History works differently from its predecessor. Instead of capturing screen images, it reads interaction events through macOS's accessibility system. OpenAI says it does not capture screenshots, screen recordings, microphone input, or system audio, and activity in private browsing mode is never recorded.

How it works

Captured events are periodically converted into text summaries and stored locally as Markdown memory files. A timeline view groups these summaries by day and time, showing which apps were involved during each session.

A core capability is detecting recurring workflows. When ChatGPT identifies a repeatable pattern—such as finding recently edited documents or generating standup summaries—the timeline suggests a "Skill." Users can then ask Codex to turn the recorded workflow into a reusable automation template, according to OpenAI.

Permissions and data handling

Access requires consent at two levels. In Business and Enterprise workspaces, an admin must first enable Computer History for the organization. That alone doesn't activate it for anyone—each individual user must also opt in and separately enable the Memories feature. Users can further control which apps and websites contribute data through include and exclude lists, and can pause or resume recording at any time from the macOS menu bar.

The feature is not yet available in the European Economic Area, Switzerland, or the United Kingdom.

On data retention, OpenAI says temporary event files are stored within the ChatGPT app group on the local Mac and deleted after 48 hours. The memory files generated from those events, however, remain as plaintext Markdown on the local file system indefinitely, until a user manually deletes them. Notably, these files are not encrypted, meaning any other program running under the same macOS user account could potentially read them.

OpenAI processes the temporary event files on its servers to generate memories but says it does not retain them afterward, except where required by law. The company states the files are not used for AI training. However, once memories surface as context in later chats, that chat content may become training data depending on a user's data control settings.

Security warnings from OpenAI itself

OpenAI's own documentation flags an elevated prompt injection risk. Because Computer History captures content from apps and websites, malicious instructions embedded in a webpage could potentially trick ChatGPT or Codex into following unintended commands. OpenAI also advises against using the feature in communication apps unless everyone in a conversation has given explicit consent, and recommends pausing recording or excluding apps that handle sensitive health, financial, or personal data.

What this means

Computer History pushes ChatGPT deeper into the operating-system layer, competing with the kind of ambient context-awareness Microsoft attempted with Recall and drew significant privacy backlash for. OpenAI's design choices—local storage, event-based rather than screenshot-based capture, dual admin/user consent, geographic exclusions for the EEA, Switzerland, and UK—reflect lessons learned from that precedent. But the unencrypted plaintext storage of memory files is a real gap: any local process with matching user permissions could read a detailed record of someone's work activity. Combined with OpenAI's own prompt injection warning, this makes Computer History a feature that offers clear productivity value—automated workflows, searchable activity history—but shifts meaningful security responsibility onto individual users and IT admins to configure exclusions correctly.