ChatGPT Connects to Apple Messages, Turning Texting Into an AI Task

Introduction

ChatGPT is moving beyond helping users write a text and toward handling parts of the conversation itself. OpenAI’s new Apple Messages plug-in connects desktop ChatGPT with Apple’s messaging application. Once authorized, users can ask the chatbot to search, sort, analyze, or edit messages. In some workflows, it can also draft and send a reply for them.

That makes the feature more than a writing aid. It places a conversational interface over a private archive of personal communications and gives the model a route to perform actions inside that archive. The convenience is clear, but so is the need to understand exactly what the integration can access and when it can act.

Key points

• Natural-language message search: Users can ask ChatGPT to locate information buried in their history or review messages received during a particular period.
• Reply assistance: The system can suggest follow-up texts based on recent conversations and prepare drafts for contacts.
• Action beyond drafting: According to the supplied information, users can ask ChatGPT to delete messages or send messages on their behalf. That creates a meaningful distinction between an assistant that proposes text and one that executes an instruction.
• Personal and professional use: The integration is also described as working with Codex and ChatGPT Work, extending the possible use cases beyond private conversations.
• Local-processing claims: OpenAI told TechCrunch that the plug-in runs locally on the user’s device and does not build a full index of all messages. The company says ChatGPT must receive a specific request before reading messages, while message content is stored locally by default rather than on OpenAI’s servers.
• Broad device permission: Setup requires Full Disk Access, a permission that allows an application to read and write files on the computer. Local storage may reduce some exposure, but it does not make the permission insignificant.
• Human review remains important: OpenAI discourages persistent approval because it removes the user’s final chance to inspect a message before it is sent.

Why it matters

The plug-in illustrates a broader shift in consumer AI. Assistants are increasingly being connected to everyday applications, where they can retrieve context and carry out multi-step tasks instead of merely generating answers. For people who manage many conversations, natural-language search and follow-up suggestions could reduce routine work. In professional settings, the same mechanism may help organize communications or prepare responses.

Messaging, however, is unusually sensitive to context. A model can misread sarcasm, relationships, implied boundaries, or the right moment to respond. Sending a message is therefore a higher-risk operation than drafting one. Deleting content introduces another irreversible action that should be handled conservatively.

Users should treat the integration as an agent with access to a private workspace, not as a harmless text editor. Limiting permissions where possible, reviewing each outgoing message, checking local storage, and avoiding persistent approval are sensible safeguards. OpenAI has provided additional explanations about local execution and retention, but the broader privacy framework still requires careful attention to product settings. The feature’s success will depend less on how many actions it can automate than on whether users can clearly control each one.

Source: TechCrunch AI