Meta’s Muse Spark rollout: new AI features, the Threads @meta.ai beta, and why users are upset

The performance claims should be read carefully. Meta says Muse Spark is competitive across areas including multimodal perception, reasoning, health and agentic tasks, but Fortune reported that Meta’s published benchmarks did not show it surpassing leading rivals across the board and would need independent validation.

Main AI features Meta is rolling out

Faster, more conversational Meta AI

Meta says Muse Spark brings faster voice responses and a more contextual assistant. In the Meta AI app, the company says users can talk naturally, interrupt the assistant, switch topics or swap languages; during a conversation, Meta AI can generate images and pull up recommendations from Reels, maps and other surfaces.

Shopping, recommendations and chat help

Meta is also using Muse Spark to add shopping and conversation-help features. RTE reported that Meta AI is introducing a shopping mode starting in the US, along with the ability to draw responses from recommendations and content shared across Instagram, Facebook and Threads.

More Meta apps and AI glasses

The model already powers the Meta AI app and website, and reports citing Meta’s announcement say it is rolling out to WhatsApp, Instagram, Facebook, Messenger and Meta’s AI glasses. Meta’s own announcement also frames smart glasses as a major use case, promising smarter AI glasses alongside faster voice in the app and new help inside conversations.

One important distinction: the Threads bot is a Meta AI product integration, while the provided Threads reports do not separately benchmark or disclose every model detail behind @meta.ai. The clearest confirmed link is that Meta is upgrading Meta AI with Muse Spark while extending Meta AI into more surfaces.

How Threads’ @meta.ai beta works

Threads is testing an official @meta.ai account that public-account users can mention in a post or reply. The bot can respond with context, recommendations or conversational answers inside the thread. Meta told TechCrunch the feature is meant to help people get real-time context about trends and breaking stories, as well as recommendations, without leaving the conversation.

Where the beta is launching first

The early Threads beta is launching first in five markets: Malaysia, Saudi Arabia, Mexico, Argentina and Singapore. Engadget reported that most Threads users cannot yet interact with the bot, but the public @meta.ai account is visible more broadly.

How @meta.ai compares with Grok on X

The Grok comparison is about the social workflow. On X, users can summon Grok in public conversations; Threads is testing a similar pattern where users mention @meta.ai to bring Meta AI into a public post or reply.

The differences are mostly about product scope and availability:

• @meta.ai is part of Meta AI and is being tested inside Threads, not X.
• The early beta is limited to five launch markets and public-account users.
• The provided sources describe the feature as Grok-like, but they do not provide an independent model-quality or accuracy comparison between @meta.ai and Grok.

Why users are criticizing the rollout

The main complaint is not that the beta exists; it is that users report they cannot block or fully avoid the @meta.ai account. Gigazine reported that the normal block option was not available for @meta.ai as of its write-up, and Engadget reported that users could not opt out or block the chatbot account even though the beta was not available to most people.

That matters because @meta.ai is designed to appear in social conversations when mentioned. If an AI account can be summoned into public replies and users cannot block it, the feature shifts from a tool someone chooses to use into something others can bring into shared spaces.

The privacy concern should be stated narrowly. The reports on this Threads beta document a user-control problem around visibility, blocking and opt-out, not a new Threads-specific disclosure that @meta.ai is reading private messages. Separately, an AI Weekly explainer on Meta AI data use says claims that Meta AI reads private messages, voice recordings and photos are not accurate, and says personalization applies to conversations with the AI chatbot rather than encrypted private messages.

Bottom line

Muse Spark is Meta’s attempt to make Meta AI faster, more multimodal and more useful across apps, smart glasses, shopping and recommendations. The Threads beta shows the next step: AI that can be called directly into public social conversations. That may make Threads more useful for fast context, but the inability to block @meta.ai is why the rollout is already becoming a trust test for Meta’s AI strategy.