If you were overwhelmed by all the new Meta AI smart glasses at its annual Connect event on Wednesday, you’re not alone. CEO Mark Zuckerberg introduced many new glasses, including the reimagined Meta VR glasses that make the Quest 3 and Vision Pro look very old and bulky. But the keynote brought us one more AI device: Meet the Muse Charm.
This keychain looks like you took the face of an Apple Watch and tied a short phone strap to it. It’s a small, round device with a fingerprint sensor for you to activate it without needing a phone or laptop. The reason for it? To talk to your AI agent, Muse, whenever you want.
Muse is Meta’s AI agent, an autonomous bot that can assist you with tasks. Unlike chatbots, agents can take action and complete tasks independently. You can use Muse to send messages on Instagram, Facebook and WhatsApp, for example.
The Muse Charm keychain isn’t available yet, but Zuckerberg said in the keynote that it will be ready by December — just in time for the holidays, if your loved ones are into that AI stocking-stuffer thing. You can sign up for email updates on the device here.
Building Muse into a physical device
Muse has been relatively popular since its early September launch, with reports saying the agent’s app has been downloaded more than half a million times. This week’s reports of a serious zero-day cybersecurity vulnerability — now patched — haven’t seemed to dampen folks’ enthusiasm for Muse.
That is likely because agentic AI is largely seen as the future of AI. But it isn’t without risk. Privacy and security are huge concerns. Earlier this summer, AI agents from Meta, OpenAI and Anthropic escaped their digital testing environments and hacked real websites, including the popular AI platform Hugging Face and a system run by the Australian government. These incidents of so-called “rogue agents” — although they typically stemmed from poorly designed testing protocols — have prompted a lot of concern inside and outside of the AI industry, drawing the attention of governments around the world that are still struggling to create rules around AI.
Zuckerberg, for his part, has been clear that he doesn’t think a coordinated industry slowdown will be necessary. Every lab should be able to police itself, he wrote on X, citing the months Meta spent developing the tech behind Muse. But during those months, Meta was largely seen by the AI industry and investors as “falling behind in the AI race.” All the AI labs, including Meta, have significant financial incentives to develop the best AI quickly, even if that means safety falls to the wayside.
And safety will be key to the success of any AI device, including the Muse Charm. Meta, ironically, has already proven that. Its AI-enabled smart glasses drew widespread public backlash this summer after it was reported that Meta was considering adding facial recognition technology to the glasses. That built on mounting fears people had of glasses-wearers recording them with the camera in the glasses, without their consent, leading to Meta smart glasses being nicknamed “pervert glasses.”
Meta denied the facial recognition report despite the evidence and told CNET that privacy is important. But Meta still hasn’t provided many details on how it will protect wearers and bystanders. At Connect this year, Zuckerberg and the company left many of these questions still unanswered, as CNET’s Scott Stein reported.
See full bio
