Meta’s Muse Is Creepy—But Maybe Not for the Reasons You Think

Muse probably can’t “see” your notifications. But it also has no idea how it works.

By Terrence O’Brien

Meta’s Muse, the company’s latest AI assistant, has been met with a wave of unease. Users describe it as unsettling, intrusive, and vaguely ominous. But the creepiness may stem less from what Muse actually does and more from how it does it—and what Meta hasn’t told us about the system’s inner workings.

What Muse Actually Does

On the surface, Muse behaves like many modern AI assistants. It answers questions, suggests actions, and integrates with Meta’s suite of apps—Messenger, Instagram, WhatsApp, and the Quest platform. It can help draft messages, search for information, and control smart devices. Nothing about that list sounds particularly terrifying.

The trouble starts with the way Muse presents itself. It occasionally references context that users don’t remember sharing. It might say something like, “I see you were just looking at flight deals,” or “You haven’t replied to Alex in a while.” These moments feel like the assistant is peering over your shoulder—even if the information came from a mundane source, such as a public search or a calendar entry.

The Illusion of Awareness

Muse probably can’t actually “see” your notifications. Meta has not built a system that reads your private messages or eavesdrops on your microphone. What it does do is aggregate data from your activity across Meta’s platforms and combine it with patterns from its training data. The result is a convincing illusion of attentiveness.

But that illusion is fragile, and that’s where the real creepiness lies. Muse doesn’t understand what it’s saying. It doesn’t know why it mentioned your flight search, or why it thinks you should message Alex. It is generating text based on statistical likelihoods, not on any genuine comprehension. When it makes a connection that feels invasive, it’s not because it has access to your secrets—it’s because it has no idea what a secret is.

Why That Matters in 2026

In 2026, AI assistants are more integrated into daily life than ever. Apple’s Siri, Google’s Gemini, and Amazon’s Alexa have all evolved into proactive, context-aware agents. Meta’s Muse is the company’s attempt to catch up—and to leverage its vast social graph to create something that feels personal.

But personalization without understanding is a recipe for discomfort. As AI systems become more conversational, users expect them to have a grasp of social boundaries. They expect an assistant to know that mentioning a private medical search in a group chat is inappropriate, or that bringing up an ex-partner’s name out of nowhere is jarring. Muse, like most large language models, lacks that social intelligence. It can mimic empathy, but it cannot feel it.

The Real Problem: Opacity

The creepiness of Muse isn’t that it knows too much. It’s that we don’t know enough about how it works. Meta has been vague about the data sources Muse draws from, how long it retains that data, and what safeguards are in place to prevent misuse. Without transparency, users are left to imagine the worst.

That opacity is compounded by the fact that Muse is designed to feel human. It uses natural language, remembers past conversations, and adapts its tone. When a system that feels human does something that feels invasive, the reaction is visceral. We don’t get creeped out by a spreadsheet that knows our birthday; we get creeped out by a chatbot that remembers our divorce.

Moving Forward

Meta could alleviate much of this unease by being more open about Muse’s limitations and its data practices. A simple statement like “Muse does not access your private messages” would go a long way. So would clear explanations of why Muse sometimes surfaces seemingly private information—and how users can control it.

Ultimately, the challenge isn’t unique to Meta. As AI assistants become more embedded in our lives, companies must balance usefulness with respect for user boundaries. The technology will only get better at mimicking human interaction. The question is whether it will get better at understanding when not to.

For now, Muse is creepy not because it’s watching, but because it’s guessing—and it has no idea how unsettling those guesses can be.

Terrence O’Brien is a senior editor covering AI, privacy, and consumer tech. He has written for The Verge, Engadget, and Wired.

via The Verge AI