Anti-Recording Arms Race as AI Wearables Loom
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Deveillance announced Spectre I in March, a hockey-puck-shaped device designed to prevent others from recording you, founded by recent college graduate Aida Baradari, who said the team is still working on its microphone-detection capability
- Apple is rumored to be developing an AI pin or pendant that would serve as an iPhone's constant eyes and ears, joining a wave of AI-enabled recording wearables that Baradari cited as the threat motivating Spectre I
- AI wearables are defeating existing jammers by deploying speech-recovery algorithms that solve the "cocktail party problem" — the same challenge DeLiang Wang at Ohio State has trained neural networks on for decades to improve hearing aids, and that Microsoft has advanced through its annual Deep Noise Suppression Challenge since 2020
- Ming Gao and a team at Nanjing University developed MicFrozen in 2023, a wearable jammer that listens to its wearer and emits ultrasonic "anti-speech" tuned to their voice plus counterfeit speech-shaped sound to mislead reconstruction algorithms
- Earlier countermeasures ranged from WWII radar chaff dropped by the RAF to swamp German screens, to 2020 work by Yuxin Chen's University of Chicago team that mounted 23 ultrasonic transducers on a single bracelet for omnidirectional jamming, to "babble tapes" of 40 layered voices in different accents that privacy law professor Woodrow Hartzog said defense attorneys once played during jailhouse meetings
- Privacy historian Finn Brunton and others argue the more durable strategy is obfuscation — feeding devices junk data, the approach behind artist Adam Harvey's anti-facial-recognition makeup and Daniel Howe and Helen Nissenbaum's TrackMeNot browser plug-in that runs decoy searches
- Hartzog warned that the defenders are outgunned: AI wearables draw on a speech-processing industry backed by billions in corporate investment, while only a few academics and small companies are building countermeasures — "we know how they usually end up for the mouse," he said
- The article drew a parallel to the Mafia's decades-long surveillance arms race with law enforcement, culminating in the FBI's 2018 Anom sting, in which 12,000 encrypted phones were sold to criminals before the feds intercepted every message
Why it matters: If AI wearables become as common as AirPods, as the article projects, ordinary conversations — workplace diplomacy, financial disclosures, romantic disclosures — could be recorded by anyone nearby without consent, and the only commercial countermeasure (Spectre I) hasn't even finished developing its microphone-detection feature. The asymmetry is stark: billions in corporate investment funds speech-recovery models, while a handful of academics and one recent graduate's startup are building the defenses.
Ask SkimNews
