Meta has spent the last two years giving its eponymous AI chatbot a prominent place in its applications and now it’s Threads’ turn. The company is starting to test a new feature that gives the Meta AI chatbot Grok-like functionality on Threads, with the ability to respond to posts with additional “context.”
To do this, Meta AI is getting an official Threads account (@meta.ai) with which users can chat alongside their other conversations in the app. The feature, which the company describes as an “early beta release,” will first roll out to Threads users in Malaysia, Saudi Arabia, Mexico, Argentina, and Singapore.
The premise sounds a lot like Grok’s original intention. Meta suggests users ask questions like “Why are people talking about the World Cup this month?” » and Meta AI will be able to respond publicly. Users can also invoke Meta AI in replies to other posts. This is a very common use case on
Of course, giving an AI chatbot that much visibility on a platform like Threads could pose some of its own challenges. On Meta has generally been more cautious about protecting his AI chatbot than X has been with Grok, who has always been presented as a sort of AI overlord. Still, it’s reasonable to wonder whether the Threads feature might be subject to the same type of user-driven manipulation that X has blamed for some of Grok’s most high-profile errors.
Meta offers, at least, a way for people to ignore the new Meta AI account on Threads. The company notes that users can choose to deactivate the account and hide replies that appear under their own posts.
The Threads test was announced as part of a broader push for the recently revamped Meta AI, which is now powered by the company’s latest Muse Spark model. In addition to the bot on Threads, Meta is testing a new way for users to chat with Meta AI about events from their group chats on WhatsApp and other apps. Just like on Threads, these “side chats,” as the company calls them, are meant to give people a way to query the bot for additional “context” based on what’s happening in the group’s thread. But, unlike Threads, these discussions take place in their own Meta AI discussion and are only visible to the person asking Meta AI to intervene, not the entire group.
Elsewhere, the new Spark model also powers a new integrated version of Live AI in the Meta AI app. Previously available only through Meta’s smart glasses, Live AI allows users to ask questions about their surroundings and get real-time responses to what their phone’s camera is pointing at. (A word of warning: Live AI was at the center of a recent report claiming that human moderators were viewing recordings of these sessions, which sometimes included intimate moments. So keep in mind that it’s possible that human eyes could see whatever you ask Meta AI.)
And, speaking of Meta glasses, the company says the Spark model is starting to roll out across its line of Ray-Ban and Oakley-branded eyewear in the U.S. and Canada, and will arrive on its display-enabled frames this summer.
Muse Spark was the first major release from Meta’s “superintelligence” group as the company attempts to lay a new foundation for its AI work in the post-Llama era. CEO Mark Zuckerberg also hinted at the company’s plans for OpenClaw-style AI agents that will run on a version of the model.