Meta unveils Muse, an artificial intelligence agent that does more than answer questions: it carries out tasks and projects on the user's behalf, from booking trips to negotiating in their name, on a dedicated, isolated virtual computer.
Meta has unveiled Muse, an artificial intelligence agent designed to carry out everyday tasks and larger-scale projects on behalf of the person using it. Unlike a conversational assistant that only answers questions, Muse takes a goal, builds a personalized plan and moves the work forward autonomously: it can open a browser, fill out forms or negotiate on the user's behalf. It keeps working even after the person closes the app and only reaches out again when it needs approval or when something relevant changes.
The system works through messaging, either in Muse's own app or on WhatsApp, and runs on Muse Spark, Meta's model built for agentic work. For payments, it relies on Link, from Stripe, with protections such as coverage for damaged or lost items and fee-free returns, plus one-time-use virtual cards that keep real bank details hidden.
Muse runs on Muse Secure VM, a dedicated, isolated virtual computer for each user that stores the agent along with its data and credentials. A second agent, called Sentinel, runs on the same machine and authorizes every connection Muse attempts to make to the internet. The agent has no access to passwords or payment methods, and checks with the person before sensitive actions such as sending an email or completing a purchase. Each user decides which apps to connect and how much access to grant, and can revoke it at any time. The company plans to launch Muse Confidential VM later this year, in which the entire virtual environment, including data and conversations, is encrypted with a key only the user holds, so that not even Meta can access it.
Muse is rolling out in the United States on iOS, Android and muse.ai, with availability on Meta's glasses expected later on. The service is free for most everyday uses and will offer paid subscription plans for those who need greater capacity.
AI assistant developed by Meta that operates through voice and text. Offers personalized responses, video editing, image generation and is available on mobile apps, web and Ray-Ban Meta ...
Assistant that carries out tasks and advances long-term goals on behalf of the user, such as booking trips, completing purchases or sending emails. Runs on a dedicated virtual machine and requests ...
06/10/2026
Mistral AI releases a preview of its largest model to date, multimodal and with open weights due at the end of the month, with notable results in ...
05/10/2026
In the coming weeks, OpenAI will add an invisible statistical signal to text generated by ChatGPT and Codex in the EU to comply with the European AI ...
03/10/2026
David Robinson, who led the writing of OpenAI's safety reports, is leaving the company and argues in an essay published in The Atlantic that the ...
30/09/2026
Google announces Gemini 4 Argon, a model that generates responses of up to one million tokens, aimed at coding, finance, legal and cybersecurity ...