ChatGPT Adds ‘Computer History’ on Mac, Logging Activity to Power New Automation Features

OpenAI has introduced a new macOS ChatGPT feature called Computer History that records a user’s app and web activity so the assistant can build a timeline of recent work, suggest automations, and help resume unfinished tasks. The company says the feature is optional, can exclude selected apps and websites, and does not collect screenshots, video, or audio.

The rollout matters because it pushes ChatGPT closer to becoming an always-aware work companion rather than a simple chatbot, while also reviving privacy concerns familiar from Microsoft’s controversial Windows Recall project.

In practical terms, Computer History gives ChatGPT and OpenAI’s coding tool Codex a way to reference what a user has been doing on their Mac. That means the assistant can answer questions such as which document was last edited, whether it was shared in Slack, or what tasks were left half-finished earlier in the day.

What OpenAI launched on macOS

OpenAI’s new Computer History feature is part of the ChatGPT desktop app for Mac. The tool turns everyday interactions into a searchable activity trail that the assistant can consult when responding to requests, helping it understand context beyond the current prompt.

The company is pitching the feature as a productivity layer for people who want ChatGPT to remember the flow of their work. Instead of starting from scratch each time, users can ask the assistant to reconstruct what happened during a work session and move directly to the next step.

How Computer History works

Computer History works by logging events from a user’s activity and organizing them into a timeline. OpenAI says the feature can then use that timeline to surface relevant details when a user asks for help, whether that means finding a recent file, summarizing a morning of work, or continuing a task that was interrupted.

In a demo highlighted by OpenAI, a member of the company’s developer experience team showed the app locating the most recently edited document, checking whether it had been shared through Slack, and producing a recap of the user’s morning.

Why the feature raises privacy questions

Computer History is designed to make ChatGPT more useful, but it also deepens the sense that AI assistants are moving closer to a detailed understanding of a user’s digital life. Even without screenshots, a timeline of clicks, app use, and keystrokes can reveal a great deal about work habits, communications, and priorities.

That is why the feature is likely to be read through the same privacy lens that shaped reactions to Windows Recall. Microsoft’s earlier feature prompted a backlash because it relied heavily on screenshots and created the impression of a near-continuous visual record of user activity.

OpenAI is trying to draw a sharp distinction. The company says Computer History does not store images, videos, or audio and instead works from “events,” which suggests a less intrusive form of logging than screenshot capture. Still, critics of AI productivity tools may argue that a timeline of events is enough to create an intimate record of user behavior.

Opt-in controls are central to OpenAI’s pitch

The most important safeguard, at least according to OpenAI’s announcement, is that Computer History is opt-in rather than opt-out. Users must actively choose to enable it, and they can also exclude specific apps and websites from being tracked.

OpenAI further says that users can delete entries if they want more granular control over what remains in the timeline. That is a meaningful feature because it gives people a way to remove mistakes, sensitive interactions, or unrelated tasks from the record the assistant can access.

Ari Weinstein, OpenAI’s product and engineering manager, said on X that Computer History will automatically ignore activity from incognito or private browser tabs.

How is this different from Windows Recall?

It is different because OpenAI says it is not taking screenshots, but the goal is similar: to help software remember what a person did so the user can ask about it later. That resemblance is likely to shape how the market, regulators, and privacy advocates interpret the feature.

Windows Recall became a flashpoint because it was framed as a memory aid for the PC, yet many users saw it as a surveillance risk. OpenAI appears to be learning from that controversy by making the feature optional and by emphasizing event-based tracking rather than visual capture.

Even so, the business logic is the same. The more an AI system knows about the user’s past actions, the better it can predict the next one. That can make the assistant feel dramatically more capable, but it also raises the stakes for trust, consent, and data handling.

What OpenAI is promising users

OpenAI’s promise is that Computer History will make ChatGPT more context-aware without crossing into the most alarming forms of surveillance. The company is presenting the feature as a productivity enhancer that helps users recover lost threads rather than as a tool for watching every move on screen.

Whether that distinction is enough will depend on how much detail the event logs reveal in practice, how transparent the controls are, and how comfortable users are with an AI assistant that knows the shape of their workday.

Why this matters for ChatGPT’s future

Computer History shows OpenAI’s broader ambition to make ChatGPT more than a conversation window. The company wants the assistant to understand user context, bridge gaps between apps, and take action across workflows that unfold over hours or days.

That shift is especially important on the desktop, where users often move between documents, messaging apps, browsers, and developer tools. An assistant that can stitch those moments together is potentially far more useful than one that only answers isolated questions.

How Codex fits into the picture

OpenAI says Codex can also reference the Computer History timeline. That matters because it suggests the feature is not only for general chat but for software development workflows, where context from prior activity can help automate coding tasks or pick up unfinished work.

For developers, the appeal is clear: fewer context switches, faster recovery after interruptions, and more seamless handoffs between brainstorming, coding, testing, and sharing. But the same benefits could make the feature especially sensitive if it records private project details or internal communications.

What users can control

OpenAI is giving users several forms of control, at least at launch. The feature must be turned on manually, and users can decide which apps and websites should never be included in the history trail.

That flexibility is important because not all activity is equally appropriate for an AI memory system. Personal banking, sensitive health information, private communication channels, and confidential work tools are all examples of places where users may want strict exclusion.

• Opt-in enrollment: users must actively enable Computer History.
• App and site exclusions: specific tools and domains can be left out.
• Deletion tools: individual entries can be removed.
• Private browsing protection: incognito tabs are ignored automatically.

What questions remain unanswered?

Several practical questions remain open. OpenAI has described Computer History at a high level, but the company has not yet provided full public detail on how long data is retained, how it is secured, whether users can export the timeline, or how much of the logged information is processed on device versus in the cloud.

Those details matter because “event” logging can still be highly revealing even if it avoids screen capture. A record of file openings, browser visits, app switching, and messaging interactions can be enough to reconstruct a user’s workflow with surprising precision.

How much context is too much?

That is the central policy question behind features like Computer History. Supporters will argue that assistants need memory to be genuinely useful, especially for knowledge workers juggling dozens of open tasks. Skeptics will respond that memory is only helpful if the user can fully see, understand, and control what the system remembers.

The answer may determine whether Computer History becomes a flagship productivity feature or another example of AI software moving faster than public comfort with surveillance-adjacent design.

Timeline of the feature’s significance

The development is notable not just as a product update, but as part of a larger evolution in consumer AI software. The category has steadily moved from text generation toward systems that observe, remember, and act.

The bigger picture for AI assistants

Computer History reflects where the AI industry is headed: toward systems that do not just answer questions, but also observe enough context to manage work on a user’s behalf. That evolution can make assistants more helpful, more sticky, and more deeply embedded in daily routines.

But the same evolution also forces companies to prove that convenience does not require unacceptable levels of data collection. Users may welcome an assistant that remembers the file they were working on yesterday. They may be much less comfortable if that same assistant becomes a detailed historian of their private digital behavior.

OpenAI is betting that careful controls and a less visual, event-based design will be enough to soften the privacy blow. The company’s success will depend on whether users see Computer History as a useful memory aid or as another reminder that AI assistants are getting uncomfortably close to the center of their lives.

For now, the launch signals that ChatGPT is moving beyond simple prompting and into the realm of continuous, context-aware assistance. That is a major product step — and one that will almost certainly keep privacy and trust at the center of the conversation.