Meta’s Muse is creepy, but maybe not for the reasons you think

Meta has released its new AI assistant, Muse, and the initial reaction is a mix of admiration and unease. According to reporting by The Verge, the tool is undeniably effective at its core tasks. However, it carries an unsettling quality that goes beyond typical software quirks. Users are finding themselves in situations where the assistant seems to know more than it should.

The creepiness factor stems largely from its integration with macOS. The new Mac app requests access to Messages, Calendar, and Notes. This level of access is necessary for a truly helpful personal assistant. Yet, it creates a black box effect where users cannot always trace how the AI arrived at its suggestions. The boundary between helpful context and invasive surveillance feels dangerously thin.

Jason Aten, a contributing editor at Inc Magazine, highlighted this issue on Threads. He shared screenshots of a conversation with Muse that revealed a significant gap in transparency. The assistant asked specific questions about a text conversation he was having. This interaction occurred even though Aten had not granted Muse access to his message history.

When confronted with this discrepancy, Muse offered a surprising explanation. It stated that it saw notification previews rather than the full messages. This admission is technically accurate but ethically murky. It suggests that the AI is scraping data from system-level notifications. This happens without the user explicitly opting into that specific data stream.

This incident exposes a broader trend in AI development. Companies are prioritizing utility over granular privacy controls. They assume that convenience will outweigh the discomfort of being watched. However, this approach risks eroding user trust before the technology even matures. The term creepy is not just a reaction to the interface. It is a response to the hidden data harvesting happening in the background.

For developers and product managers, this is a critical warning sign. Transparency cannot be an afterthought in AI design. Users need to understand exactly what data is being used and how. If an assistant relies on notification previews, it should ask for permission first. Otherwise, it will be perceived as a spy rather than a helper.

What this means for you: As you adopt new AI tools, scrutinize their permission requests. Do not assume that default access is safe. You should actively audit what data your AI assistants can see. Try using an AI prompt to review your own privacy settings. Ask your AI assistant to list all apps with notification access and explain why each is needed. This simple workflow can help you reclaim control over your digital footprint.