OpenAI has released a new tool called Computer History for the ChatGPT desktop application on the macOS operating system. This feature analyzes user activity in various programs and websites to make subsequent artificial intelligence responses more accurate and personalized. The update became available on August 13th for subscribers of the Pro, Business and Enterprise tariffs. Residents of the European Economic Area, the United Kingdom and Switzerland will have access to the innovation later, as the company plans to bring the function in full compliance with local requirements for personal data processing. Representatives of OpenAI announced the launch on the social network X, calling Computer History a logical development of an earlier experimental Chronicle project. According to the developers, the new version uses fewer tokens to maintain context and offers users much more privacy management tools.

The Computer History tool is built directly into the desktop application and runs in the background. The service records which applications and websites a person uses, and adds this data to a temporary activity feed. In subsequent dialogues, the language model can rely on this story, which significantly reduces the amount of context that the user would have to enter manually. The timeline interface allows you to return to previously completed tasks, track recurring work scenarios, and completely clear the history or delete data only for a selected period. The key condition for operation is to manually enable the option in the settings section, since data collection does not begin without this step. The function status is always displayed openly so that the user can clearly see whether tracking is underway. OpenAI describes this mechanism as a way to account for repetitive work patterns outside of the chatbot conversation itself.

In 2000, OpenAI noticeably accelerated the implementation of permanent memory in all its key products. The new release significantly reduces the computational cost of maintaining a cross-session context. On the same day, the company announced the GPT-5.6 Ultrafast mode on specialized Cerebras chips, which provides a generation rate of up to seven hundred and fifty tokens per second. At the same time, regulators in Europe and the UK are closely monitoring tools that involve permanent storage of user data. The decision to delay the launch in these regions indicates the company's desire to adapt the feature to local regulations first. Users who do not want the system to track their activity may simply not enable this feature, as it works solely on the principle of voluntary consent.

From the point of view of machine data analysis, the predecessor of the new feature contained certain technical risks. The official documentation on the Chronicle project pointed to an increased risk of injections through prompta and storing records in unencrypted form on the user's device. The focus on continuous personalization through activity analysis increases competition with other major market players such as Claude from Anthropic and Gemini from Google, which are developing similar memory mechanisms. The delayed launch in Europe can be seen as an indicator of how difficult it is to balance deep personalization and data protection requirements in different jurisdictions. The question remains whether voluntary consent will remain a sufficient barrier to user trust as the volume of data collected increases.