OpenAI has released Computer History, a macOS feature that records your clicks, keystrokes, and app switches, then converts them into a searchable timeline ChatGPT and Codex can reference. The feature replaces Chronicle, an earlier research preview that captured screenshots. Computer History captures interaction events through macOS accessibility APIs instead of what's on screen.
Private browsing activity is excluded. So are screenshots, screen recordings, microphone input, and system audio. OpenAI processes the raw events on its servers to generate text summaries, then deletes the raw data. The summaries are stored locally as plaintext Markdown files.
What does Computer History actually record?
Computer History in ChatGPT
The feature tracks interaction events: clicks, keystrokes, keyboard shortcuts, and app switches. It reads these through macOS's accessibility system, the same permissions screen readers and automation tools use. Periodically, OpenAI's servers convert these raw events into text summaries grouped by day, time, and application.
The timeline view shows these summaries chronologically. When ChatGPT detects repeating patterns in your activity, it suggests a "Skill" or automation. You can then ask Codex to build a reusable template from the recorded workflow. OpenAI cites finding recently edited documents and generating standup summaries as typical use cases.
How the permission system works
OpenAI built a multi-layered consent model. In Business and Enterprise workspaces, an admin must enable Computer History first. That alone does not activate it. Each user still has to opt in individually, and the Memories feature must be turned on.
Users control which apps and websites contribute data through include and exclude lists. Recording can be paused or resumed from the macOS menu bar. The feature is not available in the European Economic Area, Switzerland, or the United Kingdom.
Local storage, no encryption
Temporary event files sit on the Mac within the ChatGPT app group and are deleted after 48 hours. The generated memory files remain as plaintext Markdown on the local file system until you delete them manually. These files are not encrypted, so any program running under the same macOS user account could read them.
OpenAI says it does not retain the temporary event files after processing, unless required by law. The memory files are not used for AI training. But when memories appear as context in later chats, those chat contents may still become training data depending on your data control settings.
Directly relevant for users evaluating security of AI assistant accounts
OpenAI flags prompt injection risk
OpenAI's own documentation warns about 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 recommends against using the feature with communication apps unless everyone in the conversation consents. For apps handling sensitive health, financial, or personal data, the company suggests pausing recording or excluding those apps entirely.
Logicity's Take
Computer History is OpenAI's clearest move yet toward always-on ambient context for AI assistants. The shift from screenshots (Chronicle) to interaction events suggests OpenAI learned that image processing was either too slow, too resource-heavy, or too privacy-invasive. The plaintext storage and prompt injection warnings are honest, but they also shift liability to users. For product teams building on ChatGPT or Codex, this opens a path to AI that knows your workflow history without you re-explaining context every session. The tradeoff: you're trusting unencrypted local files and cloud-side processing with a detailed log of your work.
The absence of encryption for local memory files is the most glaring gap. Any malware or rogue app with user-level permissions could exfiltrate a detailed record of your activity. For enterprise deployments, this may require endpoint protection beyond what OpenAI provides.
Need Help Implementing This?
If you're evaluating Computer History for your team or building workflows around Codex automation, reach out to Logicity's expert network for implementation guidance.
Source: The Decoder / Matthias Bastian
Huma Shazia
Senior AI & Tech Writer
Produced with AI assistance and reviewed by the Logicity editorial team. Learn more in our Editorial Policy.
