OpenAI's Codex for Mac now captures screenshots and sends them to cloud servers for processing
OpenAI's Codex desktop app for Mac has added Chronicle, a feature that periodically captures screenshots, sends them to OpenAI's servers for OCR and visual analysis, then stores text summaries as unencrypted Markdown files locally. The feature requires a $100+/month ChatGPT Pro subscription and is unavailable in the EU, UK, and Switzerland.
OpenAI's Codex for Mac now captures screenshots and sends them to cloud servers for processing
OpenAI's Codex desktop app for Mac has added Chronicle, a feature that periodically captures screenshots of user activity, sends them to OpenAI's servers for processing, and stores text summaries locally as unencrypted Markdown files.
Chronicle runs background agents that capture screen content and transmit it to OpenAI servers where OCR and visual analysis generate text summaries. These summaries are saved in ~/.codex/memories_extensions/chronicle/ as Markdown files that Codex reads to understand recent user activity without explicit prompting. Raw screenshots are stored in a system temp directory and automatically deleted after six hours. According to OpenAI, screenshots are not retained on servers after processing and are not used for training.
The feature requires macOS Screen Recording and Accessibility permissions, Apple Silicon running macOS 14 or later, and a ChatGPT Pro subscription of $100 or more per month. Chronicle is unavailable in the EU, UK, and Switzerland—a geographic restriction indicating concerns about GDPR compliance around data minimization and purpose limitation.
Cloud processing vs. local-first privacy
Chronicle's architecture contrasts sharply with Microsoft Recall, which processes screenshots entirely on-device using a neural processing unit and stores them in an encrypted local database. Chronicle processes screenshots in the cloud but retains only text summaries locally—as unencrypted plain text accessible to any process on the machine.
OpenAI's documentation acknowledges specific risks: Chronicle "increases risk of prompt injection" because malicious content in captured screenshots could be interpreted as AI instructions. The memories directory "might contain sensitive information." The feature also "uses rate limits quickly," potentially constraining Pro subscribers' Codex usage. OpenAI recommends users manually pause Chronicle before meetings or when viewing sensitive content.
Greg Brockman, OpenAI's president, described Chronicle as "an experimental feature giving Codex the ability to see and have recent memory over what you see, automatically giving it full context on what you're doing."
Broader Codex transformation
Chronicle arrived in an April 16 update titled "Codex for (almost) everything" that expanded Codex from a coding assistant to a general-purpose AI workspace. New capabilities include computer use features that let Codex operate Mac apps with its own cursor, an in-app browser, image generation, persistent memory, and over 90 plugins. More than one million developers have used Codex, with usage doubling following the December launch of the GPT-5.2-Codex model.
Industry context
Screen-aware AI assistants have faced turbulent adoption. Rewind AI rebranded to Limitless before Meta acquired it in December 2025 and shut down the Mac app. Microsoft's Copilot has lost 39% of subscribers in six months, partly due to trust issues extending to Recall. A 2026 security researcher demonstration showed Recall's encrypted database could still be exploited.
Open-source alternative Screenpipe offers local-first processing with a $400 lifetime license. Perplexity's Personal Computer software turns a Mac mini into a persistent AI agent with local file access but relies on cloud processing for core intelligence.
What this means
Chronicle represents OpenAI's bet that cloud processing with temporary screenshot storage and deletion promises is sufficient to earn user trust in ambient AI. The choice prioritizes utility and server-side capability over the local-first privacy architecture competitors have adopted. The EU/UK/Switzerland exclusion and unencrypted local storage suggest OpenAI recognizes regulatory and security risks but is proceeding anyway in markets where it can. Success depends entirely on whether users trust OpenAI's data handling promises and whether those promises hold as the feature scales beyond research preview to general availability.
Related Articles
OpenAI Publishes GPT-6 Astra Prompting Guide With Banned 'Slop Words' List
OpenAI has published detailed prompting guidance for GPT-6 Astra, addressing the model's tendency to over-clarify, over-test, and use clichéd AI phrasing. The documentation includes specific prompts to encourage more autonomous action and a blocklist of banned words and phrases.
OpenAI Launches GPT-6 Astra With Half the Message Allowance of GPT-5.6 Sol
OpenAI has begun rolling out GPT-6 Astra to top-tier ChatGPT plans, the API, Azure, and AWS Bedrock. The model delivers roughly half the usage allowance of GPT-5.6 Sol across comparable plans, with Plus and Business users gaining access in the coming days.
Simon Willison's Pelican Benchmark Shows GPT-6 Astra Outperforming GPT-5.6 Sol at Every Reasoning Level
Developer Simon Willison ran his signature 'pelican riding a bicycle' SVG test on newly-accessed GPT-6 Astra across five reasoning levels, comparing results against GPT-5.6 Sol, Terra, and Luna. Even Astra's lowest reasoning setting reportedly beat every Sol output, though Astra costs roughly twice as much per token.
OpenAI Lists GPT-6 Astra Pro on OpenRouter: Same Model, Higher-Compute Reasoning Mode
GPT-6 Astra Pro, now listed on OpenRouter, is the existing GPT-6 Astra model configured to run with reasoning.mode set to 'pro' for higher-quality output on complex tasks. It carries a 1M-token context window and tiered pricing from $5/$25 to $20/$100 per million input/output tokens depending on the serving tier.
Comments
Loading...