OpenAI is adding a new capability to the ChatGPT desktop app on macOS that might remind you of Microsoft’s controversial Recall feature—but with a key difference: no screenshots. Called Computer History, the opt-in tool tracks your clicks and keystrokes to build a timeline of your activities, which ChatGPT and Codex can then reference when you ask for help. The goal is to make the AI assistant more context-aware, allowing it to suggest automations, pick up tasks you’ve left half-done, and even recap how you spent your morning.
What Exactly Does Computer History Do?
At its core, Computer History is a local activity log. When enabled, it records the applications you use, the documents you edit, and the websites you visit—at least on the surface. OpenAI describes this as a timeline of “events” rather than a continuous recording of your screen. That distinction matters because it supposedly avoids the heavy privacy baggage associated with Windows Recall’s screenshot-based approach.
In a demo shared by OpenAI, a developer named Dominik Kundel shows the feature in action. The ChatGPT app looks up the last document he edited, checks whether it was shared with colleagues via Slack, and provides a concise recap of how he spent the morning. This kind of contextual awareness could be a major shift in how people interact with AI assistants. Instead of manually feeding ChatGPT information about your current task, the app already knows what you’ve been working on and can act on that understanding.
The feature is designed to work with Codex, OpenAI’s agentic coding tool, as well. By referencing the timeline of your recent activity, Codex could theoretically pick up a coding task you started hours ago, understand what you were trying to do, and help you finish it. Similarly, ChatGPT could suggest automations based on patterns it detects in your behavior—for instance, if you regularly open the same set of files every morning, it might offer to create a shortcut that opens them all at once.
How It Differs From Windows Recall
When Microsoft introduced Windows Recall in 2024, it immediately drew criticism from privacy advocates and security researchers. The feature worked by taking periodic screenshots of everything on your screen, then making that content searchable using AI. While Microsoft argued that all data was processed locally, the mere existence of a screenshot database on your computer was enough to worry many users. Security researchers even found that the database was not encrypted and could be accessed by malware.
OpenAI’s Computer History takes a different route. Instead of screenshots, it relies on what the company calls “events.” These events are likely metadata about user interactions—things like window titles, app names, timestamps, maybe even keyboard input—but not images, videos, or audio. OpenAI explicitly states that Computer History does not capture those forms of media. This makes the feature less invasive in one sense, but it still involves recording a detailed log of your computing activity, which can be just as revealing in the wrong hands.
The absence of screenshots also means Computer History can’t recreate a pixel-perfect view of your screen. But it can still tell a story about your workday. An event log showing that you opened a confidential spreadsheet, then switched to a messaging app, then visited a competitor’s website is enough to paint a picture. For anyone dealing with sensitive information, that kind of trail is a legitimate concern.
Privacy and Controls: What OpenAI Is Doing
OpenAI has made Computer History opt-in rather than opt-out. That means you have to actively turn it on in the ChatGPT desktop app’s settings. The company is also offering granular controls. You can exclude specific applications and websites from being tracked, and you can delete individual entries or clear the entire history at any time. This is a more user-friendly approach than Windows Recall, which was initially criticized for being on by default and for making its controls difficult to access.
Ari Weinstein, a product and engineering manager at OpenAI, added on X that Computer History will automatically ignore content in incognito or private browser tabs. That’s a thoughtful touch, as it prevents accidental recording of sensitive browsing sessions. However, users should be aware that not all web traffic may be covered by this safeguard—it depends on how the event tracking integrates with different browsers and apps.
Despite these safeguards, the feature is still “kind of creepy,” as some observers have pointed out. Even if the data stays on your device, the idea that your AI assistant knows what you did on your computer hour by hour is a big adjustment. OpenAI claims the timeline is used to improve your personalized experience and to train models, but for many people, the thought of an AI system logging their keystrokes will be hard to swallow.
The Training Data Question
One of the most significant aspects of Computer History is its potential role in training AI models. In the original article, OpenAI frames the feature as turning your actions into training data. That means your clicks and keystrokes could be used to teach ChatGPT how humans interact with software, which could improve future versions of the model. But this raises important questions about consent, data retention, and the potential for misuse.
If you’re using ChatGPT for work, your Computer History might include proprietary company data, client details, or internal communications. Even if the data is anonymized, the mere fact that it’s being collected and used for training could violate corporate security policies. Employers may need to decide whether they want ChatGPT’s desktop app installed on company machines at all, or at least enforce strict settings that disable Computer History.
OpenAI has not yet published a detailed data retention policy specific to Computer History. It’s unclear how long the timeline is stored locally, whether it’s synced to OpenAI’s servers, and whether users can permanently delete all traces of it. The company says you can delete entries, but without transparency about backend processes, users are left to trust OpenAI’s assurances. Given the history of AI companies facing scrutiny over data handling, that trust is not guaranteed.
Why This Matters for AI Assistants
Computer History is part of a broader trend toward “agentic” AI—systems that can take actions on your behalf rather than just respond to queries. For an AI to be truly useful as an agent, it needs context. It needs to know what you were doing, what you care about, and what you’ve already tried. A timeline of your computer activity is one way to provide that context.
Microsoft’s Recall, Apple’s upcoming on-device AI, and now OpenAI’s Computer History are all moving in the same direction. They all require some level of persistent memory about user activity. The challenge is balancing utility with privacy. Users might love a feature that automatically summarizes their entire morning or finds a document they edited days ago—but they also don’t want their every move recorded and potentially exposed.
OpenAI’s approach of using “events” instead of screenshots is a deliberate compromise. It allows for a rich understanding of workflows while avoiding the privacy nightmare of a visual recording. Whether that compromise is enough to reassure users remains to be seen. The feature is new, and real-world testing will reveal not only how useful it is, but also what unintended consequences might emerge.
For now, Computer History is available only on macOS, but it’s likely to expand to Windows and possibly other platforms in the future. As the feature matures, OpenAI will need to continue refining its privacy controls and provide clearer explanations of how the data is used. In a world where digital privacy is more important than ever, the success of such a feature will depend as much on trust as on technical capability.
Source:The Verge News