OpenAI
ChatGPT on Mac Can Now Read and Reply to iMessages With New Plugin Amid Apple-OpenAI Legal Fight

OpenAI has released a new plugin that lets ChatGPT read, search and send messages through Apple's Messages app on Mac, an integration that arrives at a notably tense moment in the relationship between the two technology companies, according to multiple outlets covering the rollout.

The Apple Messages plugin, released Aug. 20, allows ChatGPT to interact with iMessage, SMS and RCS conversations stored on a Mac, enabling users to search past messages, summarize conversations, draft replies, and send messages directly through the app on their behalf. According to Dataconomy, the integration is available across all ChatGPT subscription tiers, including free accounts, and functions within the ChatGPT Work and Codex modes of the macOS desktop application.

The feature carries specific hardware and software requirements. According to MacRumors, the plugin requires an Apple silicon Mac and does not support older Intel-based machines. Users must also grant several system-level permissions for the integration to function, including macOS Full Disk Access, along with access to contact names and automation controls. OpenAI has published example prompts to demonstrate the plugin's intended use, including one that instructs the assistant to "check my calendar and reply to [name] with a few times I'm free for dinner next week."

According to 9to5Mac, OpenAI built the integration using established macOS system technologies, specifically AppleScript and the operating system's Accessibility settings, the same underlying tools that power other ChatGPT desktop features such as its broader Computer Use capability. Archyde similarly confirmed that the plugin relies on sanctioned, officially documented Apple automation interfaces rather than any reverse-engineered method of accessing Apple's messaging protocols.

Privacy protections built into the feature center on a default approval requirement before any message is sent. According to Business Standard, ChatGPT asks users to confirm both the content of a proposed message and its intended recipients before it is actually transmitted, giving users a final opportunity to review outgoing communications. Users do have the option to grant standing, persistent approval for a specific chat, allowing future messages in that conversation to be sent without repeated confirmation prompts. However, OpenAI has explicitly cautioned against doing so. According to 9to5Mac, the company's plugin guide warns that persistent approval "removes the final opportunity to review a message before it is sent," and recommends that users retain per-message approval, particularly for conversations that could potentially contain misleading or untrusted instructions.

OpenAI has emphasized that the plugin was designed with privacy considerations in mind, particularly around how message data is processed. According to Dataconomy, the company said the plugin processes data locally on the user's device and avoids creating any general, standing index of all of a user's messages. Engadget's coverage of the rollout similarly noted that the feature operates in an opt-in manner, requiring users to actively install and configure the plugin through the ChatGPT Plugins interface, where it is listed under the platform's Public category, rather than being enabled automatically for existing users.

The integration also documents a known technical limitation involving certain automation configurations. According to Implicator.ai, OpenAI has acknowledged that tasks configured with "Full access" permissions, which disable standard approval prompts, can prevent the Apple Messages plugin from displaying the confirmation screen it needs before sending a message. OpenAI's guidance for resolving that issue instructs affected users to "switch to Ask for approval or Approve for me and try again."

The plugin's release comes at a particularly complicated moment in the broader relationship between OpenAI and Apple. According to Archyde and MacRumors, Apple has filed a lawsuit against OpenAI alleging theft of trade secrets, with Apple's filing claiming that OpenAI deliberately targeted and hired away Apple employees specifically to obtain confidential corporate information. Neither company has publicly confirmed the extent to which that ongoing legal dispute directly involves or is separate from the newly released Messages integration, and Apple has not issued any public statement specifically addressing OpenAI's new iMessage-related plugin, according to Blockonomi.

The move also revives longstanding questions about third-party access to Apple's Messages ecosystem, a topic that has drawn scrutiny in the past. Archyde noted that in 2024, Apple repeatedly disabled the third-party chat app Beeper Mini's access to iMessage until the app's developers ultimately abandoned efforts to restore that integration, an episode widely viewed at the time as evidence of Apple's historically cautious approach toward allowing outside applications to interact directly with its messaging platform. Because OpenAI's new plugin relies on sanctioned macOS automation frameworks rather than any workaround targeting Apple's underlying messaging protocols directly, the situation differs technically from the earlier Beeper dispute, though Archyde noted that the current legal battle between the two companies adds an additional layer of regulatory and operational uncertainty to the feature's long-term viability on Apple's platform.

The Apple Messages plugin is not the only recent example of OpenAI expanding ChatGPT's access to sensitive, device-level personal data on Apple hardware. According to Blockonomi, OpenAI recently introduced a separate capability allowing ChatGPT to access Apple Health data on iPhone and iPad devices, representing another category of highly personal information the company has moved to integrate into its assistant's capabilities. Blockonomi characterized the Messages plugin as part of a broader, ongoing pattern in which OpenAI has steadily expanded ChatGPT's reach into private data repositories across Apple's broader device ecosystem.

Consumer privacy advocates have raised concerns regarding third-party AI systems gaining access to private communications data, even when such access is governed by permission-based controls, according to Blockonomi's reporting. OpenAI has maintained that its layered, multi-step authorization process, requiring explicit system permissions alongside default message-by-message send approval, adequately addresses those underlying privacy concerns.

As of this report, Apple has not issued a formal public statement addressing OpenAI's newly released Apple Messages plugin. With the feature currently limited to Apple silicon Macs running ChatGPT Work or Codex modes, and unavailable within standard ChatGPT chats, on the web, on mobile devices, or through the Codex command-line interface, users interested in trying the integration will need to specifically opt in through the ChatGPT desktop app's Plugins section, granting the required system permissions before the assistant can begin reading, searching or sending messages through Apple's native Messages application on their behalf.