OpenAI rolls out Computer History to Pro, Business and Enterprise; Europe to wait

OpenAI is launching the Computer History feature in the desktop version of ChatGPT for Pro, Business and Enterprise users, while users in the European Economic Area, the United Kingdom and Switzerland will have to wait a few weeks. In Business accounts the admin must enable the feature at the workspace level before any employee can opt in; Pro users can enable it themselves. The default setting is off for everyone.

Computer History turns activity in apps and websites into a sequence of memories and a timeline that ChatGPT and Codex can retrieve as context. Users can ask natural-language questions such as “What was I working on before the interruption?” or “Where is the document I searched for this morning?”, receive a daily stand-up summary, or ask the model to identify recurring processes and convert them into skills or Codex automations. When a file, Slack thread or Google Doc is relevant, the model uses the history to locate the source and reads it directly.

Unlike the earlier research prototype Chronicle, which relied on screenshots, Computer History is rebuilt to record only interaction events—keystrokes, clicks, keyboard shortcuts, app switches and the data macOS exposes through its accessibility framework. It does not capture screen images, video, audio or activity in private browsing windows. Events are periodically translated into text summaries stored locally. In settings users can view a timeline filtered by day and hour, select which apps and sites participate, pause collection from the menu bar, and delete individual items or the entire history at any time.

Permissions are tiered: admins in Business and Enterprise environments must approve access in the permissions settings, after which each team member can opt in personally. The feature also requires Memories to be enabled and is unavailable when connecting via an API key or through Amazon Bedrock. For organizations that rely on API integrations, the feature simply does not exist, a limitation that must be considered in internal rollout plans.

OpenAI did not give a specific date for availability in the blocked regions beyond “the next few weeks”, nor did it address future availability for Free or Plus users. At present the tool is limited to the higher-paying tiers, with a hard dependency on Memories and admin permissions, meaning organizations will need to decide whether the benefit of “picking up where I left off” outweighs the administrative overhead and the local data exposure the feature entails.