ChatGPT’s Computer History tracks your clicks and keystrokes
The Verge
•2026-08-16T10:56:40-04:00
📰 What Happened
OpenAI's ChatGPT desktop app on macOS has added a feature called Computer History that turns a user's activity into training data. It records clicks and keystrokes, builds a timeline of how a person works, and can suggest automations or pick up tasks the user left half-finished. ChatGPT and Codex can reference the timeline when processing requests.
🔍 The Backstory
The feature resembles Windows Recall, Microsoft's controversial tool for recording on-screen activity, though ChatGPT's version captures action-based data rather than screenshots. Watchdog groups and privacy advocates have long worried about AI assistants harvesting intimate details of how people work. OpenAI appears to be moving toward more agentic helpers that act on a user's behalf, which requires deeper access to behaviour.
🎯 Why It Matters
The tool offers genuine convenience by automating repetitive work, but it deepens the trade-off between usefulness and privacy. It shows how far AI assistants will go to observe user behaviour, and raises questions about consent, data storage and who shrinks the line between a helpful helper and an always-on watcher. Worth watching for how users and regulators respond.
OpenAI's ChatGPT desktop app on macOS has added a feature called Computer History that turns a user's activity into training data. It records clicks and keystrokes, builds a timeline of how a person works, and can suggest automations or pick up tasks the user left half-finished. ChatGPT and Codex can reference the timeline when processing requests.
The feature resembles Windows Recall, Microsoft's controversial tool for recording on-screen activity, though ChatGPT's version captures action-based data rather than screenshots. Watchdog groups and privacy advocates have long worried about AI assistants harvesting intimate details of how people work. OpenAI appears to be moving toward more agentic helpers that act on a user's behalf, which requires deeper access to behaviour.
The tool offers genuine convenience by automating repetitive work, but it deepens the trade-off between usefulness and privacy. It shows how far AI assistants will go to observe user behaviour, and raises questions about consent, data storage and who shrinks the line between a helpful helper and an always-on watcher. Worth watching for how users and regulators respond.