ChatGPT’s Computer History tracks your clicks and keystrokes — a new data-timeline for assistants
The Verge AI reports a MacOS desktop feature that records user actions to build a training and reference timeline for ChatGPT and Codex. This capability promises more contextualized automation and smoother task completion by recalling prior requests and patterns. On the flip side, it raises questions about data provenance, collection scope, and user consent for training data. In enterprise contexts, the ability to auto-suggest actions and reference historical activity could dramatically improve productivity—provided users have clear visibility into what data is captured, how it’s stored, and how long it’s retained.
From a product perspective, the feature reflects a broader move toward personal AI agents that learn from ongoing user behavior. For developers, it emphasizes the need for robust opt-in controls and privacy safeguards, especially in sensitive environments where highly personal data might be processed. Regulators will likely scrutinize the boundaries of user data usage, the transparency of data pipelines, and the ways in which users can review or delete their traces. In the immediate term, this capability could become a differentiator for ChatGPT on macOS, enabling deeper automation and more intuitive workflows, while also challenging teams to craft governance models that keep user data handling aligned with evolving privacy expectations.
As AI assistants deepen their integration into daily work, stakeholders should observe performance trade-offs, data minimization principles, and the governance frameworks that govern how much the system remembers and why. The real success metric will be whether users feel more empowered without sacrificing trust or control over their own information, a balance that remains at the heart of sustainable AI adoption.