OpenAI's Computer History tracks user activity across apps and websites, turning it into a searchable timeline with memories that ChatGPT and Codex can use as context.
The feature replaces the earlier research preview called "Chronicle", which relied on screenshots. Computer History works very differently, recording interaction events instead of capturing what's on screen. That means clicks, keystrokes, keyboard shortcuts, and app switches, all read through macOS's accessibility system. Screenshots, screen recordings, microphone input, and system audio aren't captured, and activity in private browsing mode is never included, OpenAI says.
The captured events are periodically converted into text summaries and stored locally as Markdown memory files. The timeline view groups these summaries by day and time and shows which apps were involved.
Repeated workflows turn into reusable skills
A key part of Computer History is its ability to spot recurring workflows. When ChatGPT detects repeatable patterns in a user's activity, the timeline suggests a "Skill" or automation. Users can then ask Codex to build a reusable template from the recorded workflow.
Typical use cases include finding recently edited documents or generating standup summaries, according to OpenAI. ChatGPT can pull from the recorded activity to locate the right source and access it directly.
Admins and users both have to opt in
OpenAI built a multi-layered permission system for Computer History. In Business and Enterprise workspaces, an admin has to explicitly enable the feature first. That alone doesn't activate it, though, since each user still has to consent individually. The Memories feature needs to be turned on too.
Users control which apps and websites contribute data through include and exclude lists, and recording can be paused or resumed at any time from the macOS menu bar. The feature isn't available yet in the European Economic Area, Switzerland, or the United Kingdom.
Memory files are stored locally but not encrypted
Temporary event files sit on the Mac within the ChatGPT app group and get deleted after 48 hours, according to OpenAI. The memory files generated from them remain as plaintext Markdown on the local file system until the user manually deletes them.
These files aren't encrypted, meaning any program running under the same macOS user account could potentially read them. To generate the memories, OpenAI processes the temporary event files on its servers but says it doesn't keep them after processing, unless required by law.
The files aren't used for AI training, OpenAI says. But when memories show up as context in later chats, those chat contents may still become training data depending on the user's data control settings.
OpenAI itself warns about prompt injection risks
OpenAI's own documentation flags a heightened prompt injection risk. Because Computer History captures content from apps and websites, malicious instructions embedded on a webpage could trick ChatGPT or Codex into following them.
OpenAI also warns against using the feature with communication apps unless everyone in the conversation has given explicit consent. For apps that handle sensitive health, financial, or personal data, OpenAI recommends pausing recording or excluding those apps entirely.