OpenAI has rolled out Computer History, a macOS feature that tracks user activity across apps and websites and turns it into a searchable timeline with memories that ChatGPT and Codex can use as context. The feature, reported in detail by The Decoder, marks the company's most aggressive push yet into ambient, always-on personal context — and it arrives with a privacy rulebook users will want to read closely.
Computer History replaces an earlier research preview called Chronicle, which relied on periodic screenshots to build its record of user activity. The new version works very differently: instead of capturing what is on screen, it records interaction events — clicks, keystrokes, keyboard shortcuts, and app switches — all read through macOS's accessibility system, according to OpenAI's documentation. For more context on this story, see our ongoing breaking AI news.
How the timeline works
The 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 and shows which apps were involved, effectively giving ChatGPT a searchable memory of what you did on your Mac and when you did it.
OpenAI says the feature deliberately avoids the most invasive data types: screenshots, screen recordings, microphone input, and system audio are never captured, and activity in private browsing mode is never included. Typical use cases the company cites include finding recently edited documents and generating standup summaries, with ChatGPT able to pull from recorded activity to locate the right source file and access it directly.
Repeated workflows become Skills
A key part of the feature is pattern detection. When ChatGPT detects repeatable workflows in a user's activity, the timeline suggests a Skill or automation, and users can ask Codex to build a reusable template from the recorded workflow. That turns raw activity logs into something closer to a personal automation engine — the system does not just remember what you did, it learns how you work and offers to productize it.
Permissions and availability
OpenAI built a multi-layered permission system for Computer History. In Business and Enterprise workspaces, an administrator has to explicitly enable the feature first — and that alone does not activate it, since each user still has to consent individually, with the Memories feature also switched on.
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. Availability is geographically restricted: the feature is not available in the European Economic Area, Switzerland, or the United Kingdom, a carve-out that mirrors the regulatory distance OpenAI has maintained with several recent releases.
The privacy fine print
The documentation details carry caveats that privacy-conscious users should weigh. Temporary event files sit on the Mac within the ChatGPT app group and are deleted after 48 hours, but the memory files generated from them remain as plaintext Markdown on the local file system until the user manually deletes them. Those files are not 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, and the company says it does not keep them after processing unless required by law. The files are not used for AI training, OpenAI says — but when memories surface as context in later chats, those chat contents may still become training data depending on the user's data control settings.
OpenAI's own injection warning
Most notably, OpenAI's documentation itself flags a heightened prompt injection risk. Because Computer History captures content from apps and websites, malicious instructions embedded on a webpage could theoretically trick ChatGPT or Codex into following them. The company also warns against using the feature with communication apps unless everyone in the conversation has explicitly consented, and recommends pausing recording or excluding apps entirely for software that handles sensitive health, financial, or personal data.
The race to own personal context
The launch positions OpenAI in the emerging contest to build assistants that understand not just what users ask, but everything they do. Where chatbots once answered discrete questions, products like Computer History point toward assistants with continuous awareness of a user's work — able to reference the spreadsheet from Tuesday, the codebase from this morning, or the recurring weekly report without being told.
That ambition explains both the feature's appeal and its friction. The same always-on record that makes ChatGPT genuinely useful — an unbroken, searchable history of your digital work — is precisely what makes it a sensitive asset, stored unencrypted on disk and processed on OpenAI's servers before being summarized locally. The company's decision to keep Europeans out of the feature for now suggests it expects regulators to see the same tension.
For now, the feature is opt-in at every layer, granular in its controls, and explicit about its limits — a more careful rollout than the name might suggest. But the direction is unmistakable: the next battlefield in consumer AI is not better answers to better prompts. It is memory, and Computer History is OpenAI's bid to make the Mac itself the source of it.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
