OpenAI has expanded access to its macOS interaction logging feature, ‘Computer History,’ to the UK, EEA, and Switzerland, raising privacy and security questions about the collection and storage of sensitive user data.
OpenAI has widened access to its ChatGPT “Computer History” feature, bringing the macOS recording tool to the European Economic Area, Switzerland and the UK from 20 August. The change matters because the feature had launched with those markets excluded, despite OpenAI positioning it as a way for ChatGPT and Codex to build a searchable record of past activity on a Mac. According to Notebookcheck and TechRepublic, the feature remains opt-in and is disabled by default.
Computer History works by logging interaction events on macOS, including clicks, keystrokes and app switching, then turning that material into summaries and a timeline that ChatGPT can query later. TechRepublic reported that it relies on ChatGPT Memories and allows users to restrict which apps and websites feed into the record. Private browsing is excluded, and OpenAI says it does not capture screenshots, microphone input or system audio.
OpenAI’s own warnings are more notable than the feature’s convenience. Notebookcheck says the company cautions that sensitive material may be stored without encryption, while other programmes running under the same macOS user account may be able to access the generated files. Temporary event data is deleted after 48 hours, but the resulting memory files remain on the local machine until the user removes them.
The company also advises users to switch the feature off during conversations with other people unless they have consent, because chats, emails and video calls may be pulled into the summary. Another concern is prompt injection, where hidden instructions embedded in websites or applications could influence the model. Notebookcheck says OpenAI accepts that the feature increases that risk, while TechRepublic notes that users can pause collection from the Mac menu bar and delete individual entries or longer sections of history.
What happens to the data after processing is more nuanced. Notebookcheck reports that the temporary event files are sent to OpenAI for summarisation and are not retained or used for training once processed. However, if a generated memory is later used in a chat, that chat content may feed model improvement depending on the user’s privacy settings in ChatGPT.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
