OpenAI Computer History: ChatGPT's New Memory Feature

OpenAI has launched "Computer History," a new feature that transforms your digital activity into a searchable memory timeline for ChatGPT and Codex. This system tracks clicks, keystrokes, and app switches across various applications and websites, using macOS's accessibility system. Here's the thing: it replaces an earlier preview that used screenshots. Computer History records interaction events, not screen content, and it specifically avoids capturing screenshots, audio, or activity in private browsing. What's interesting is how it works: captured events are converted into text summaries and stored locally as Markdown memory files. These files are not encrypted. OpenAI states it processes temporary event files on its servers but doesn't keep them unless legally required, and these files are not used for AI training. However, if memories appear in chat contexts, those chat contents might become training data depending on user settings. A key benefit is the ability to spot recurring workflows. Computer History can suggest "Skills" or automations from repeatable patterns, letting Codex build reusable templates. Users can control which apps and websites contribute data, and recording can be paused. The bottom line is that both admins and individual users must opt-in, and the feature is not yet available in the European Economic Area, Switzerland, or the UK. This gives users a new way to leverage their past activity for AI assistance, but with important considerations for data privacy and security.