ChatGPT Adds Apple Messages Plugin, Letting Users Draft and Send Texts

OpenAI has introduced a new Apple Messages plugin for ChatGPT that lets users connect their Messages inbox and use the chatbot to sort, analyze, draft, edit and even send texts on their behalf. The feature could make ChatGPT more useful for everyday communication, but it also raises fresh questions about privacy, message access and how much control users should give an AI assistant.

The rollout marks another step in OpenAI’s push to turn ChatGPT into a broader productivity tool rather than just a conversational interface. It also expands the chatbot’s reach into one of the most personal corners of digital life: private messaging.

What OpenAI just launched

OpenAI has added support for Apple Messages through a new plugin that allows ChatGPT to interact with a user’s text-message history and assist with replies. The company says the integration is designed to help people find, organize and act on messages more efficiently, whether they are using ChatGPT for personal tasks or through its work-focused products.

In practical terms, the plugin can do more than surface old conversations. It can help draft responses, suggest follow-ups, search for details buried in message threads and, if the user allows it, send texts directly.

A promotional demo for the feature shows a user asking ChatGPT to recommend follow-up messages based on conversations received the previous day, underscoring the tool’s role as an active writing assistant rather than a passive search tool.

How does the Apple Messages plugin work?

The plugin works by connecting ChatGPT to Apple Messages so the assistant can operate on the contents of a user’s inbox. OpenAI says the feature can be used to sort, analyze, edit and search messages, and it also works with Codex and ChatGPT Work, which suggests the company sees message handling as relevant to both consumer and professional use cases.

What users can do with it

According to OpenAI’s description, the plugin supports a range of common messaging tasks:

• drafting replies to incoming texts
• suggesting follow-up messages
• sending texts on the user’s behalf
• deleting messages
• searching for information in message history
• organizing or analyzing conversations

That combination makes the tool more ambitious than a simple summarizer. It moves ChatGPT into the realm of a communications operator, with permission to interpret a user’s message history and take action based on prompts.

Why this matters for ChatGPT users

The integration matters because it positions ChatGPT as a front end for a core phone function that many people rely on all day. Text messaging is often the default channel for family coordination, scheduling, customer communication and workplace logistics, so any AI that can write or send texts becomes immediately useful — and potentially risky.

How OpenAI is framing privacy and control

OpenAI is presenting the plugin as a locally run feature, saying it operates on a user’s device and does not create a complete index of message history. That claim is meant to address the most obvious privacy concern: whether a chatbot could quietly ingest and store large amounts of sensitive personal data.

But the company’s explanation leaves room for interpretation. “Runs locally” and “doesn’t create an index” may reassure some users, yet those phrases do not fully explain what data is accessed, how long it remains available to the model, or what guardrails exist around sensitive conversations.

OpenAI has said the plugin runs locally on the user’s machine and does not build a full index of messages, but the company has not yet fully spelled out the technical details behind that assurance.

OpenAI has also urged users to remain attentive when ChatGPT is allowed to send messages. The company warns against enabling persistent approval, a setting that removes the final manual checkpoint before a text is sent. In OpenAI’s view, that last review step is an important safeguard that helps prevent accidental or inappropriate messages.

The company cautions that turning on persistent approval eliminates the user’s last opportunity to review a message before ChatGPT sends it as if it were the user.

Why is this raising privacy concerns?

This is raising privacy concerns because messaging apps contain some of the most sensitive information on a phone, including personal relationships, financial arrangements, location details and work-related discussion. Giving an AI assistant the power to read, summarize and send texts introduces new questions about consent, data retention and misuse.

Even if the assistant is operating locally, a system that can access message history still needs robust permission design. Users will want to know whether ChatGPT can see entire threads, how it treats attachments, whether deleted messages remain accessible in any form and how much of the interaction is retained for model improvement or debugging.

What remains unclear

OpenAI’s initial explanation does not answer every question. Among the details that still matter are:

1. what exact permissions users must grant to connect Messages
2. whether the plugin can access all message threads or only selected ones
3. how the system handles sensitive content such as codes, addresses or private photos
4. what happens if ChatGPT misreads a message and sends the wrong reply
5. how thoroughly activity is logged for the user to review later

Those questions are important not only for privacy advocates, but also for ordinary users who may be tempted by convenience without realizing how much access they are authorizing.

How does this fit into OpenAI’s bigger strategy?

This fits into OpenAI’s broader strategy of embedding ChatGPT into daily digital workflows. The company has spent the past several years turning its chatbot into a tool that can search, write, summarize, code and manage tasks rather than merely answer questions in a chat window.

By moving into messaging, OpenAI is extending that strategy into the communications layer of the smartphone — a move that could make ChatGPT more indispensable while also making it more invasive.

The ability to help with private messages also creates a new bridge between consumer AI and professional productivity software. OpenAI’s mention of Codex and ChatGPT Work indicates the company does not want users to think of the plugin as a novelty; it wants it to become part of everyday work communication, too.

Why Apple Messages is a strategic target

Apple Messages is a strategic target because it is deeply embedded in the Apple ecosystem and widely used for personal communication, group coordination and business exchanges. An integration there can be especially sticky: if ChatGPT becomes the easiest way to search and reply to messages, users may begin relying on it repeatedly throughout the day.

That kind of stickiness is valuable for OpenAI because it deepens engagement and broadens the chatbot’s utility. It also increases the importance of trust, since any assistant handling messages must be seen as reliable, discreet and reversible when mistakes happen.

What the feature could mean for everyday users

For many people, the feature may simply feel like a smarter autocomplete system. Someone who is busy can ask ChatGPT to draft a response to a delayed friend, summarize a long thread from a colleague or surface the location shared in a family conversation from days earlier.

For others, the convenience may be less appealing. A chatbot that can access texts may feel too close to personal correspondence, especially if it can also delete or send messages. That tension between convenience and intrusion is likely to shape how widely the feature is used.

Likely use cases

• catching up on long group threads
• finding a phone number, address or appointment detail
• drafting professional follow-up texts
• keeping conversations moving when the user is unavailable
• rewriting a message to sound more polished or concise

These are straightforward tasks, but the appeal of the plugin may lie in how much time it saves across dozens of small interactions. Messaging is repetitive by nature, and that makes it fertile ground for AI assistance.

What could go wrong?

What could go wrong is that an assistant with sending privileges can make a small mistake feel much bigger. A misplaced word, an unintended recipient, a tone-deaf reply or a deleted message could create embarrassment or operational problems in both personal and business settings.

There is also a broader trust issue. The more an AI system acts on behalf of a user, the more people must trust its interpretation of intent. In messaging, intent matters enormously, because texts often involve nuance, urgency and social context that AI can misunderstand.

Risk areas to watch

• accidental sending without adequate review
• misinterpretation of context or tone
• overbroad access to sensitive message history
• uncertainty over local processing and retention
• user overreliance on automated replies

OpenAI’s warning about persistent approval is telling. It suggests the company recognizes that the final approval step is more than a usability feature; it is a major trust boundary.

How this compares with other AI assistant trends

This is part of a broader industry pattern in which AI tools are becoming more agent-like. Instead of simply generating text, assistants are increasingly being asked to retrieve information, edit content, take actions and connect directly to apps people already use.

In that sense, the Apple Messages plugin is not just an isolated product update. It is an example of how AI companies are trying to move from advice to action. The goal is to make the model useful enough that users can delegate routine digital chores with minimal friction.

That shift is likely to define the next phase of consumer AI. A model that can write a paragraph is helpful. A model that can draft, decide and send a text is much closer to becoming a digital assistant in the traditional sense.

Timeline of the rollout

The new plugin arrived on August 20, 2026, as part of OpenAI’s latest wave of product expansion. Here is a simplified timeline of the feature’s emergence and immediate implications.

What happens next?

What happens next will depend on how users react once the novelty wears off. If the plugin proves reliable and easy to control, it could become one of the more practical examples of ChatGPT’s expanding utility. If the privacy story remains murky or if users worry about automated sending, adoption may stay limited to cautious early adopters.

OpenAI has not yet fully clarified every technical detail, and TechCrunch said it reached out for further information. That suggests the launch may be the first step in a longer conversation about how much access an AI assistant should have to private communications.

For now, the central takeaway is clear: ChatGPT is no longer just helping people write messages from scratch. It is increasingly being positioned to manage the messaging workflow itself — from finding old conversations to sending new ones — and that makes the product both more useful and more sensitive than before.

As the feature rolls out, the real test will not just be whether it works. It will be whether users are comfortable letting an AI stand so close to their most personal and immediate form of digital communication.