ChatGPT macOS Computer History logs draw infostealer warnings

ChatGPT’s macOS Computer History stores unencrypted activity summaries that infostealers could steal; OpenAI says it is off by default.

OpenAI’s Computer History feature in the ChatGPT macOS app logs user actions for persistent context and stores AI-summarized activity as unencrypted plain-text Markdown files. Those files are readable by other processes on the same user account, so macOS infostealers can collect a concise record of a victim’s digital activity. Cyber Security News specifically warns that stealers such as AMOS and MacSync could harvest the summaries and also flags higher prompt-injection risk. GBHackers describes the same readable Markdown memory files and says they could support social engineering and extortion, without naming those stealers or the default setting. The sources do not contradict each other; the first is more specific, including OpenAI’s acknowledgment that the feature is off by default.

• OpenAI’s ChatGPT Computer History feature for the macOS desktop app logs user actions and saves AI-summarized activity as unencrypted plain-text Markdown files.
• Those local memory files are accessible to other processes on the same Mac user account.
• Cyber Security News names macOS infostealers AMOS and MacSync as likely to harvest the pre-summarized activity.
• The same report says the feature also raises prompt-injection risk; GBHackers says stolen summaries could aid social engineering and extortion.
• OpenAI has acknowledged the risk and says the feature is off by default.
• Both reports are dated 2026-09-23 (Cyber Security News, then GBHackers).