OpenAI Adds Apple Messages Plugin to ChatGPT for Mac
OpenAI has added an Apple Messages plugin to ChatGPT for Mac, allowing the AI to read, search, and send iMessage, SMS, and RCS conversations. The feature is available across all plans but currently limited to Apple silicon Macs.
OpenAI has released a new Apple Messages plugin for ChatGPT on macOS, enabling the assistant to read, search, and send messages directly from a user's Mac. The integration works across iMessage, SMS, and RCS, according to OpenAI's release notes.
What the plugin does
The Apple Messages plugin lets ChatGPT read and search Messages conversations on a Mac and prepare or send new messages on the user's behalf through the Messages app. OpenAI says the feature is available on all subscription plans within the ChatGPT desktop app for macOS, and it also works inside ChatGPT Work and Codex.
The plugin does not enable remote interaction with ChatGPT through Messages, and it does not function in standard ChatGPT chat conversations — it is scoped specifically to the desktop app's plugin system. For now, it only works on Apple silicon Macs; Intel-based Macs are not supported.
Approval and privacy controls
Given the sensitivity of granting an AI model access to personal message history, OpenAI says sending is gated by user approval by default — ChatGPT will only send a message after the user confirms both the message content and its recipients.
OpenAI's plugin guide, referenced in the release notes, outlines risks associated with granting persistent approval, steps for revoking plugin access, and a known issue in which certain automated tasks can disable approval prompts. OpenAI has not disclosed further technical details on how message data is processed or stored, and no independent security review of the plugin has been published.
Other changes in this release
The same update bundles several unrelated features:
- Pinned chat sync: Pinned conversations now stay synchronized between the ChatGPT desktop app and iOS.
- Codex thread snapshots: Users can share a read-only, point-in-time snapshot of a local Codex thread. Snapshots from personal accounts are accessible to anyone with the link; workspace snapshots are restricted to workspace members. OpenAI says Codex redacts known secret patterns but recommends manual review before sharing, since sensitive data could remain.
- Site co-editing: Workspace members can now be invited to co-edit a published ChatGPT-hosted Site, with owners retaining control over access, settings, analytics, and version rollback.
- Custom Site URLs: Some Site owners can change their hosted URL without redeploying; the previous URL redirects automatically.
- Computer History expansion: A feature introduced last week is expanding to Pro, Business, and Enterprise users in the EEA, Switzerland, and the UK. It requires Memories to be enabled and remains off by default.
No pricing changes accompanied this update.
What this means
This is a product feature addition to the existing ChatGPT desktop app, not a new model release — it changes how users interact with ChatGPT on macOS rather than altering any underlying model weights. The integration signals OpenAI's continued push toward agentic, OS-level assistants that can act on a user's behalf across native apps, following a similar pattern to browser and file-system integrations rolled out earlier in 2026.
The approval-gating and known bypass issue in automated tasks are worth watching: any AI system with read/write access to personal communications carries real risk if permission controls fail. Apple silicon-only support also suggests the plugin relies on system-level APIs not uniformly available on older Intel Macs, which may delay broader rollout.
Related Articles
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.
Gemini Overlay on Android Adds Minimize Button for Multitasking Bubble
Google is widely rolling out a new Minimize button for the Gemini overlay on Android, which collapses conversations into a floating bubble users can drag or tap to expand. The feature currently supports six fixed positions and has a reported bug that resets bubble placement after each minimization.
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.
OpenAI's GPT-6 Astra Cuts Hallucinations, But Indirect Prompt Injection Attacks Still Succeed 8.5% of the Time
OpenAI's new GPT-6 Astra model shows major improvements in hallucination rates and jailbreak resistance over predecessor GPT-5.6 Sol, according to OpenAI's system card. However, indirect prompt injection attacks hidden in documents still succeed 8.5% of the time in external testing by Gray Swan, down from 27% but still above rival Claude Opus 5's 4.8% rate.
Comments
Loading...