OpenAI’s Chronicle feature lets Codex read your screen, raising privacy concerns

OpenAI’s Chronicle captures recent screen activity to build contextual memories so the Codex coding assistant can interpret references, identify relevant sources, and understand users’ tools and workflows without repeated prompts. The opt-in macOS research preview for ChatGPT Pro stores temporary screen captures locally, processes selected frames through ephemeral Codex sessions, and includes pause controls and privacy considerations such as prompt injection risk and unencrypted local markdown storage. #Chronicle #Codex

Keypoints

Chronicle captures recent screen activity to build memories that give Codex contextual understanding.
Codex can use Chronicle to identify and read files, Slack threads, Google Docs, dashboards, or pull requests as primary sources.
The feature is an opt-in research preview for ChatGPT Pro on macOS and requires Screen Recording and Accessibility permissions.
Screen captures are stored temporarily on the user’s computer, processed through ephemeral sessions, and deleted after six hours.
Users can pause Chronicle for sensitive content, but memories are saved as unencrypted markdown and may affect rate limits and injection risk.