> ## Content Index > Fetch the complete content index at: https://www.implicator.ai/llms.txt > Use this file to discover other available public pages before exploring further. # OpenAI Lets ChatGPT Read and Send iMessage, SMS and RCS on Apple Silicon Macs - URL: https://www.implicator.ai/chatgpt-apple-messages-imessage-sms-rcs-mac/ - Published: 2026-08-21T15:18:22.000Z - Updated: 2026-08-21T15:18:22.000Z - Description: OpenAI’s new Apple Messages plugin lets ChatGPT read, search, draft and send iMessage, SMS and RCS conversations on Apple-silicon Macs. Every send requires approval by default, but users can grant persistent permission to a specific chat. OpenAI also documents a Full access problem that can prevent - Author: Marcus Schuler - Tags: AI News OpenAI released an [Apple Messages plugin](https://learn.chatgpt.com/docs/plugins?surface=app&ref=implicator.ai#app-use-apple-messages-from-codex) on [Aug. 20, 2026](https://learn.chatgpt.com/docs/changelog?ref=implicator.ai) that lets ChatGPT read conversations on a Mac and prepare or send replies through Apple’s Messages app. Users can grant the assistant standing permission to send later messages in a specific chat without another approval prompt, though every send requires approval by default. The plugin works with iMessage, SMS and RCS conversations already present on the computer by using local macOS tools. What Changed - ChatGPT’s Apple Messages plugin can read and search local conversations and prepare or send iMessage, SMS and RCS replies on Apple-silicon Macs. - The plugin is included with every ChatGPT plan but works through ChatGPT Work and Codex, not regular ChatGPT chats, web, mobile, Intel Macs, Codex CLI or the IDE extension. - Sending requires approval of the message and recipients by default; users can separately grant persistent sending permission to one Messages chat. - A documented Full access issue can prevent the required confirmation from appearing, leaving the requested message unsent rather than silently bypassing review. AI-generated summary, reviewed by an editor. [More on our AI guidelines](https://www.implicator.ai/about/). ## Where the plugin works OpenAI includes the plugin with every ChatGPT plan, but the initial release is limited to the Apple silicon version of its macOS desktop app. The feature is usable from ChatGPT Work and Codex inside that app. It is not available in ordinary ChatGPT chats, on the web or mobile, through Codex CLI, or in the IDE extension. Intel Macs are also excluded from this release. Users install it from the desktop app’s Plugins tab, then start a new ChatGPT Work or Codex conversation and grant the requested macOS permissions. Messages cannot be used as a remote control for ChatGPT. The plugin works with conversations that are already available on the Mac, so sending a text to the computer does not start a ChatGPT session. ## What ChatGPT can read and do Once installed, the plugin can search local message history, summarize conversations, draft replies and send messages through the Messages app. OpenAI’s demonstration showed a user asking ChatGPT to find conversations missed the previous day and suggest follow-up messages before sending a reply. OpenAI says the capability runs on the Mac and uses operating-system tools including AppleScript and Accessibility. It requires the user’s consent and does not create an index of all messages. The setup screen tells users that ChatGPT will access their on-device Messages history, and macOS requires [Full Disk Access](https://support.apple.com/en-us/guide/mac-help/mchl211c911f/mac?ref=implicator.ai), Contacts and Automation. Together, those permissions allow ChatGPT to find recipients, read the local archive and operate the Messages app. No independent security audit of the plugin is documented. OpenAI’s description says the work runs locally and no message index is created, but it does not fully explain which message content, if any, may reach OpenAI’s network while a request is processed. Apple has not commented on the integration. At SXSW in March 2025, Signal President Meredith Whittaker warned of profound privacy and security implications from agentic AI. She compared the approach to “[putting your brain in a jar](https://techcrunch.com/2025/03/07/signal-president-meredith-whittaker-calls-out-agentic-ai-as-having-profound-security-and-privacy-issues/?ref=implicator.ai),” but was discussing agentic AI broadly, not this Apple Messages plugin. FREE WEEKDAY MORNING BRIEFING Don’t miss the next AI story that matters. The Implicator Morning Briefing filters the AI news cycle to the stories worth your attention and explains their consequences. From San Francisco, every weekday at 4:45 a.m. Pacific, 7:45 a.m. Eastern. Email address Send me tomorrow’s briefing Check your inbox. Click the link to confirm. About five minutes. No hype. No spam. ## Approval can persist By default, ChatGPT shows the proposed message and its recipients before sending, and the user must approve both. “Allow once” applies to that send. Selecting “Always allow sending to this chat” lets ChatGPT send later messages to the same conversation without another approval prompt. The permission is tied to that Messages chat, not every conversation in the archive. Know someone who'd find this useful? [✉️ Email it to a friend in one click](mailto:?subject=A%20newsletter%20I%20think%20you%27d%20like&body=This%20is%20one%20of%20maybe%20three%20newsletters%20I%20actually%20read.%20The%20rest%20just%20pile%20up%2C%20unread%2C%20judging%20me.%0A%0AAnd%20yes%2C%20this%20email%20mostly%20wrote%20itself%2C%20which%20is%20a%20little%20on%20the%20nose%20for%20an%20AI%20newsletter.%20Doesn%27t%20make%20it%20wrong.%20implicator.ai%20is%20good.%0A%0ASubscribe%20free%3A%20https%3A%2F%2Fwww.implicator.ai%2Fsubscribe%2F%3Futm%5Fsource%3Dnewsletter%26utm%5Fmedium%3Dforward%26utm%5Fcampaign%3Demail%5Fforward), or they can [subscribe free here](https://www.implicator.ai/subscribe/?utm%5Fsource=newsletter&utm%5Fmedium=forward&utm%5Fcampaign=forward%5Fto%5Fcolleague). OpenAI tells users to retain per-send approval for conversations that may contain untrusted or misleading instructions. Its guide warns that persistent approval “removes your final chance to review a message before ChatGPT sends it as you.” Users can revoke a chat’s persistent sending permission under Settings > Computer use > Messages. Administrators of managed workspaces can separately disable Apple Messages through the existing Computer Use control. ## When confirmation cannot appear OpenAI also documents a known issue involving tasks configured for Full access or otherwise set to disable approval prompts. In that state, Apple Messages may be unable to display the confirmation required to send, leaving the requested message unsent. The known issue prevents the required confirmation and is not documented as a silent send that bypasses review. The prescribed fix is to change the task’s access mode and retry. OpenAI’s instruction is: “Switch to Ask for approval or Approve for me and try again.” Frequently Asked Questions What can ChatGPT do with Apple Messages on a Mac? The plugin can read and search conversations already present on the Mac, summarize them, draft replies and send messages through Apple’s Messages app. It supports iMessage, SMS and RCS. Which Macs and ChatGPT accounts support the Messages plugin? OpenAI includes it with every ChatGPT plan, but this release requires an Apple-silicon Mac. It works from ChatGPT Work and Codex in the macOS desktop app, not regular ChatGPT chats, Intel Macs, web, mobile, Codex CLI or the IDE extension. Can ChatGPT send an Apple Message without approval? Every send requires approval of the proposed message and recipients by default. A user can explicitly grant persistent sending permission to a specific Messages chat, which removes later per-send prompts for that conversation. What permissions does the Apple Messages plugin require? The setup gives ChatGPT access to on-device Messages history and requires macOS permissions for Full Disk Access, Contacts and Automation so it can find recipients, read the local archive and operate Messages. Does the documented Full access issue silently send messages? No. OpenAI says Full access or another setting that disables approval prompts may prevent Messages from showing the confirmation needed to send, leaving the requested message unsent. OpenAI recommends switching to Ask for approval or Approve for me and retrying. AI-generated summary, reviewed by an editor. [More on our AI guidelines](https://www.implicator.ai/about/). [Apple Leased Its Mind to GoogleApple rebuilt Siri on Foundation Models it co-developed with Google's Gemini, and said its most demanding requests will run on Google's cloud servers and Nvidia chips. The privacy company now depends The Implicator![](https://www.implicator.ai/content/images/2026/06/2026-06-08-22.51.16-apple_gemini_core@2x.webp)](https://www.implicator.ai/apple-leased-its-mind-to-google/) [Anthropic Adds Passport Checks to Claude After Privacy-Driven User SurgeAnthropic has begun asking some Claude users to verify their identity with a government photo ID and, in some cases, a live selfie, according to a new company help page that says the checks apply to sThe Implicator![](https://www.implicator.ai/content/images/2026/04/2026-04-16-03.50.13-anthropic_adds_passport@2x.webp)](https://www.implicator.ai/anthropic-adds-passport-checks-to-claude-after-privacy-driven-user-surge/) [Moltbook Was Broken, Fake, and Brilliant. Meta Paid Anyway.On January 31, two days after Matt Schlicht's AI assistant finished building a social network for robots, security researchers at Wiz found the front door wide open. No locks. No alarms. 1.5 million AThe Implicator![](https://www.implicator.ai/content/images/2026/03/2026-03-10-10.17.29-robot_phonebook@2x-1.webp)](https://www.implicator.ai/moltbook-was-broken-fake-and-brilliant-meta-paid-anyway/)