OpenAI now lets people text Dot, its always-on AI agent in ChatGPT, from the ordinary messaging app on their phone. The catch is in OpenAI’s own small print: the texts pass through an outside company on the way, so they are not end-to-end encrypted, even over iMessage. For now it costs at least $100 a month to try.
What’s new
A Dot is a personal AI agent that keeps working on tasks in the background, such as tracking a project or watching for an email, and comes back when it needs a decision. Until now you reached it in ChatGPT, Slack or Microsoft Teams. From Wednesday, OpenAI’s ChatGPT release notes say Pro users in the US can also text it.
According to OpenAI’s setup guide, you open your Dot’s profile, choose Texting, verify your phone number and save the contact card it sends. It works over iMessage on iPhones and RCS, the newer messaging standard, on Android phones. You can send text, photos and voice notes, and the Dot can text you first when something changes. If it wants to do something that needs your approval, it sends a link back into ChatGPT. Group chats, calls and FaceTime are not supported.
Who can use it
Only ChatGPT Pro subscribers in the US, as a beta. Pro now comes in three tiers at $100, $200 and $500 a month. Dots themselves are not offered on Pro plans in the European Economic Area, the UK or Switzerland, so readers there can’t use this at all for now.
The privacy trade-off
OpenAI says your Dot can only text you, you’re the only one who can text it, and setting it up doesn’t give it or OpenAI access to your other conversations. But the messages are carried by Linq, a third-party messaging company. As Windows Report quoted from OpenAI’s support page: “Because Linq processes these messages on the way to OpenAI, the conversation is not end-to-end encrypted between you and your dot, even when you use iMessage or RCS.”
In plain terms, end-to-end encryption means only the sender and the receiver can read a message. Here, a company in the middle handles the content before it reaches OpenAI. OpenAI says it ran extensive security testing with Linq. Turning texting off disconnects your number, but it doesn’t delete past chats or what the Dot has saved to memory.
Why it matters
AI assistants are moving into the places people already talk. Apple has been giving Siri memory of past chats on the iPhone, and now OpenAI wants its agent in the same Messages app as your friends and family. But a text to your Dot is not as private as a text to a person, so it is worth thinking twice before sending it passwords, health details or anything you wouldn’t put in an email.
The post You Can Now Text ChatGPT’s Dot Agent, but It Isn’t End-to-End Encrypted appeared first on DailyAI.
