Meta has introduced Muse, a personal AI agent designed to automate digital tasks, from booking travel to selling cars, with a strong emphasis on security and privacy. The agent, which can be accessed through a dedicated app, website, or WhatsApp, uses a model called Muse Spark, described as Meta's most capable model to date. Muse is initially available in the United States on iOS, Android, and the web, with support for Meta's AI glasses to follow.
The company stresses that none of the data shared with Muse will be used for advertising purposes, a crucial distinction given Meta's history with user data. Payments are handled through Stripe's Link, which provides purchase protections, and Meta has implemented a security architecture that includes isolated virtual machines for each user and a system called Sentinel to monitor and control actions taken by the agent.
Muse is seen as Meta's attempt to compete with other AI agents like OpenClaw and Instinct, and its development is part of Meta Superintelligence Labs' efforts to advance AI technology. The agent's capabilities include making suggestions, remembering preferences, and turning saved content into actionable lists. Meta's chief AI officer, Alexandr Wang, has emphasized the company's vision for personal superintelligence that assists people in achieving their goals.
The launch of Muse comes with a free tier and subscription plans at $20 and $100 a month, according to Axios. Wang has managed expectations, suggesting that most users will find the free tier sufficient for their needs. The elaborate safety features, including Sentinel, are designed to prevent potential mistakes, such as filling out wrong forms or making incorrect purchases, by requiring human approval for critical actions.
Meta is also planning to release a confidential version of Muse later this year, which will run in a trusted execution environment, allowing users to manage their access keys locally and ensuring that no one, including Meta, can access the user's agent VM without permission. This move is part of Meta's collaboration with Moxie Marlinspike, known for developing the end-to-end encrypted messaging app Signal and the privacy-focused AI platform Confer.
The introduction of Muse and its focus on privacy and security mark a significant step for Meta, given the company's past challenges with user trust. By emphasizing the security and privacy features of Muse, Meta aims to convince users that the agent is a safe and reliable tool for automating digital tasks.
Cet article a été rédigé avec l'assistance de l'IA.
News Factory APP - actualités agentiques pour booster votre SEO et AEO.