The news
OpenAI expanded ChatGPT's memory tools on macOS with a new Computer History option. The feature tracks user activity in other applications and on websites. It then feeds that record back into ChatGPT so the model can produce responses that reference recent work without the user restating details.
Context
ChatGPT already stored some user-supplied facts under its existing memory settings. The new capability moves beyond those manual entries by pulling data directly from the computer itself. The change affects only the macOS version of the ChatGPT application at this stage. No equivalent function has been announced for other platforms.
Details
The Computer History feature records which apps a user opens and which sites they visit. It passes that log to ChatGPT so the assistant can reference recent tasks when answering questions. OpenAI states the goal is to deliver responses that already contain relevant context from the user's current workflow. The announcement gives no timeline for wider rollout or for similar tools on Windows or mobile.
Privacy and control questions
The single source provides no information on data retention periods, encryption in transit, or whether the recorded activity stays on the device for any period before upload. It also does not state whether users can view the collected log, pause collection per application, or delete entries after they have been sent. Existing memory controls in ChatGPT allow users to turn memory on or off and to delete specific stored items, but it remains unclear whether Computer History inherits those same toggles or operates under a separate set of rules.
Why it matters
Engineers who keep several applications open at once often repeat the same background information to chat-based tools. A system that reads activity across those windows can reduce that repetition, but it also creates a new data surface inside the model. Users must decide whether the convenience of automatic context outweighs the fact that their app and browser history now leaves the device and enters OpenAI's systems. Because the feature lives only on macOS for now, teams that standardize on other operating systems will see no change in their daily tooling. Over time the same approach could appear elsewhere, shifting the default expectation that language models stay ignorant of everything happening outside their own chat window. Organizations that treat activity logs as sensitive will need to review whether the added context justifies the added exposure. The absence of published retention or deletion details means any evaluation must currently rest on general OpenAI privacy policies rather than feature-specific guarantees. For individual developers the trade-off is straightforward: faster answers about the file just edited or the documentation page just opened, at the cost of a persistent record of that work. For teams the calculation includes audit requirements, data residency rules, and the possibility that activity data could later be used to train models unless explicit opt-outs are confirmed. The limited initial release on macOS gives OpenAI time to gather feedback before committing the same mechanism to Windows or mobile clients, where usage patterns and regulatory exposure differ. Until more specifics appear on data handling and user controls, the feature remains an experiment whose long-term value depends on whether the context it supplies proves more useful than the visibility it removes.
---
Sources:
{"word_count": 682, "sources_used": 1}
No comments yet