AI Wearables vs. Jammers: The Coming Surveillance Arms Race

Soon the person next to you at a bar might be wearing a microphone and camera disguised as glasses, a necklace or a lapel pin. An essay in The Atlantic examines the arms race now forming between AI-enabled wearables and the countermeasures meant to defeat them.

AI wearables head for the mainstream

Today's AI recorders work as silent note-takers, personal assistants or memory aids, and the article argues they may soon stop being niche. Apple, which runs the world's largest personal-tech ecosystem, is rumoured to be developing an AI pin or pendant that would act as an iPhone's constant eyes and ears; AI accessories, it suggests, could one day be as common as AirPods.

In March, the start-up Deveillance said it is building Spectre I, a hockey-puck-shaped device that purports to stop others from recording you. Founder Aida Baradari, a recent graduate alarmed by the surge of AI recorders, would not explain how the jammer's signals work beyond saying they resemble speech; a promised microphone-detection feature, she said, is still being developed.

Ultrasonic jammers meet speech-recovery AI

Jamming has history. Early gadgets blasted white noise; newer ones emit inaudible ultrasonic frequencies, exploiting a quirk of microphone hardware. In 2020, a University of Chicago team led by Yuxin Chen mounted 23 ultrasonic transducers on a single bracelet so the signal could be steered.

AI, however, can strip that noise back out. Advanced pins and glasses use speech-recovery algorithms to isolate one voice among many - the "cocktail party problem". DeLiang Wang of Ohio State University has spent decades training hearing-aid neural networks to recognise the frequencies and rhythms of speech; the newest models can infer missing syllables.

Fighting data with data

Some researchers prefer obfuscation to silencing. Finn Brunton of UC Davis points to Adam Harvey's anti-facial-recognition clothing and makeup, and to the TrackMeNot extension, which buries real searches among decoy queries. Law professor Woodrow Hartzog recalls defence lawyers playing "babble tapes" layered with 40 tracks of voices. In 2023, a team led by Ming Gao, now at Nanjing University, demonstrated a jammer worn by a speaker that emits ultrasonic "anti-speech" tuned to their voice, plus counterfeit speech-shaped sound.

An unequal race

Wearables may always have the edge: they draw on a speech-processing industry that takes in billions, while a few academics and small firms build defences. Future models might skip audio altogether and read lips, as HAL 9000 does in 2001: A Space Odyssey, or reconstruct speech from the vibrations of a glass of water. "The thing about cat-and-mouse games is that we know how they usually end up for the mouse," Hartzog says.