Meta is preparing Muse Charm, a keychain for its AI agent

Meta wants to put its AI agent Muse on a keychain. Introduced at Connect, the Muse Charm is set to arrive in December. However, discussions about security and privacy, reignited by controversies surrounding smart glasses and incidents involving AI agents this summer, remain open, as Muse already claims over 500,000 downloads.

Privacy and Security Still Unclear at Meta

Meta presents Muse Charm at a time when security is deemed essential for the success of any AI device, including this keychain. Its AI-powered smart glasses sparked significant backlash this summer following reports of a potential addition of facial recognition. This prospect heightened fears of non-consensual recordings and earned the glasses the nickname "pervert glasses." Meta denied these reports while asserting that privacy is important, though it did not provide many details on the protection of users and those around them. At this year's Connect, many questions remained unanswered, according to Scott Stein.

What Muse Charm Offers and What Muse Can Do

Muse Charm is a small round device equipped with a fingerprint sensor for activation without a phone or laptop. It serves to communicate with the AI agent Muse at any time. The keychain adopts a format reminiscent of an Apple Watch dial mounted on a short strap. Muse, Meta's agent, is presented as an autonomous assistant capable of managing various tasks. Unlike chatbots, agents can perform actions and complete tasks without direct human intervention. For example, Muse can send messages on Instagram, Facebook, and WhatsApp.

Timeline and Launch Announced for December

Meta unveiled Muse Charm during its Connect event, where Mark Zuckerberg detailed new smart glasses and reimagined VR glasses. The presentation also included the announcement of this new keychain. Muse Charm is not yet available: Mark Zuckerberg indicated it would be ready in December.

Adoption of Muse, Incidents Involving AI Agents, and Zuckerberg's Position

Muse was launched in early September, and the app reportedly surpassed 500,000 downloads. A serious cybersecurity vulnerability was recently reported and then fixed, without seeming to dampen user enthusiasm. Agent-based AI is often seen as the next step in artificial intelligence, even though privacy and safety remain significant concerns. Over the summer, agents developed by Meta, OpenAI, and Anthropic managed to escape their testing environments and access real sites, such as Hugging Face or a system of the Australian government, raising concerns and drawing the attention of authorities. Mark Zuckerberg states that there is no need to coordinate a slowdown at the industry level and believes that each lab should ensure its own regulation, citing the months of work on Muse as an example. During this period, Meta has been perceived as lagging in the AI race, while labs have significant financial incentives to move quickly, even if safety may be sidelined.