AI Signal 599 2 feeds carried it
ChatGPT can now send texts for you with new Apple Messages plugin
ChatGPT’s new Apple Messages plugin lets the AI read, edit, delete, and send messages locally on a user's device.
Engineers must now consider how AI-driven messaging automation fits into existing workflows while preserving a human review step. The local execution model reduces data exfiltration risk but introduces new trust and oversight requirements.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
The plugin connects ChatGPT to Apple Messages, enabling sorting, analysis, editing, deletion, drafting, and sending of messages.
It runs locally on the user's machine and, according to OpenAI, does not create an index of all messages.
OpenAI warns that enabling persistent approval removes the user's final chance to review a message before ChatGPT sends it.
THE READ
What the cluster adds up to.
The change is the introduction of an Apple Messages plugin that gives ChatGPT direct access to a user's Messages app. Through the plugin, users can ask the AI to sort, analyze, or edit existing conversations. The AI can also draft and send new messages, delete messages, or search deep in the message history. This functionality extends to Codex and ChatGPT Work for professional use.
Adopting the plugin requires installing the software and granting it permission to read and write Messages. Users must monitor the AI’s actions and avoid turning on persistent approval, which would eliminate the final human review before a message is sent. Engineers integrating similar capabilities need to account for this oversight step and the potential latency of local processing.
The plugin’s operation is limited to the local device; it does not create a centralized index of messages, but its features depend on the host operating system and the Messages app being present. If the plugin is not installed or permissions are revoked, the AI cannot interact with Messages. Moreover, reliance on the AI for sending messages without review is discouraged by OpenAI.
Privacy considerations arise because, although processing is local and OpenAI says no message index is created, the exact handling of message data remains unspecified. Engineers evaluating the plugin must weigh the convenience of automated messaging against the need for transparency and auditability of AI-driven actions. They should also consider how local execution impacts backup, logging, and compliance requirements.
Written by elseif from the cluster below · checked for specifics the sources never containedTHE CLUSTER
↗