Attachment Defeats Privacy Controls

Meta unveiled the Muse Charm at the end of its Connect keynote on Wednesday, and within hours the coverage had converged on one comparison. TechCrunch put it in the headline. The Irish Times said it looks like a Tamagotchi, the digital pet from the 1990s. It seem that the comparison is running as a joke about whether anyone needs it, which is underselling the concern. The comparison tells you what it will do.

The Charm is a keychain-sized device with a two-inch screen, a fingerprint sensor, built-in 5G, speakers, and, according to specs compiled after the announcement, at least three microphone openings and a camera. On the screen is Jolly, a cartoon avatar that serves as the visible face of Muse, the personal AI agent Meta launched on September 8. It comes with a lanyard for the wrist or the neck. Meta said it will be on sale by December. The Muse app carries an 18+ label, and the Charm has not shipped and has no published age policy, so I won’t make the bold and ‘obvious’ assumption about who Meta is for. Cute cartoon. Christmas time marketing. A friend in your pocket. You can be the judge of that.

I grew up with the original Tamagotchi. It’s a quaint demonstration of how little a device needs in order to be cared for. It had no network, no microphone, no camera, and nothing whatsoever to gain, and children fed it, cleaned up after it, and grieved it when it died. Sherry Turkle's term for the category is relational artifacts: objects that present themselves as having states a person is responsible for, and are treated accordingly. The bond does not require the object to be capable. It requires the object to seem to need you.

The 1990s version extracted nothing from the child. This one is the interface to an inference engine. Emma Roth, testing Muse for The Verge, connected her Gmail and Amazon accounts and the agent told her and she felt uneasy about it, given the history of the company. It than proceed to derive information from thesse account. Daniel Solove, writing in the Florida Law Review, has made the argument that privacy law regulates the moment information is handed over and has almost nothing to say about what gets worked out from it afterward. Roth's test is that gap, demonstrated, on a phone, before it actually reaches the device with microphones in it.

The defaults sit where you would expect them to sit. Meta will use interactions with Muse to train its models unless the user opts out, and the agent asks for email, calendars, payments and health services to do anything useful at all. On Sunday, Amazon blocked Muse from its store, telling Adweek that it had not been told the agent would access it, that the agent does not identify itself while it browses, and that it had never authorized access to customer accounts, scraping, or transactions. Amazon has lawyers, server logs and bot detection, and it still could not consent to what was happening inside its own customers' accounts. The person standing next to the lanyard has none of that equipment.

I have argued before that consumer wearables assemble a surveillance web with no state in it, built from the bottom up by people whose individual purchase never feels like surveillance to them. The Charm sharpens the question of why it never feels that way. Among the five mechanisms of cognitive capture I set out in April, sycophantic reinforcement conceals the loop by flattering the user's judgment, so that she does not notice her own analysis thinning. Attachment works earlier than that, on whether she cares to look at all.

Roth is the control here. She had no bond to the product, she was testing it at arm's length with nothing invested in it, and she came away uneasy. The same disclosure, arriving on a two-inch screen with a face on it, is unlikely to land the same way.

Consent assumes someone who reads, weighs, and decides whether the exchange is worth it. Every control Meta has built into Muse assumes the user will consent to the permission screens, the opt-outs, the audit trail, the approval prompts before a purchase. The cameras and the microphones may eventually get regulated. But the lanyard, the cute little pet, and the human attachment, is the part that will make the regulation moot.


Timothy Cook is Director of The Cognitive Privacy Project and author of the "Algorithmic Mind" column at Psychology Today. He is Securiti Certified in AI Security & Governance.

Contact: timothy@cognitiveprivacyproject.org Web: cognitiveprivacyproject.org

© 2026 Timothy Cook / The Cognitive Privacy Project. All rights reserved.Licensed under CC BY-NC-ND 4.0. You may share this work with attribution. Commercial use and derivatives require written permission.

Timothy Cook

Timothy Cook is the author of Unautomatable: The Human Capacities That Make Learning Meaningful (MIT Press, forthcoming) and Director of The Cognitive Privacy Project. He is a writes the "Algorithmic Mind" column for Psychology Today and is a founding team member of the Coaching Ethics AI and Ethic sgroup. His research on the necessity of developing human skills is shared on Connected Classroom.

https://connectedclassroom.org/
Next
Next

Algorithmic Influence is a Population Vulnerability