A Privacy “Aura” for the Age of AI Wearables? Why Ultrasonic Jammers Keep Running Into Physics

Background

Once upon a time, “always listening” meant the smart speaker on your kitchen counter. Now it includes earbuds waiting for a wake word, smart glasses with a live assistant, lapel pins that transcribe everything you say, and phones that hop into voice mode the moment you whisper a command. The line between a personal assistant and a roving microphone keeps getting blurrier, and that has triggered predictable pushback: technical countermeasures that promise to reclaim a bubble of privacy.

For cameras, that backlash has included reflective glasses, infrared LED hats, and face-paint patterns meant to defeat recognition. For microphones, the go-to counter is ultrasound: high-frequency sound above human hearing that can overwhelm nearby mics without blasting audible noise. The concept isn’t new—researchers and artists have built “bracelets of silence,” desk pucks, and acoustic cloaks for years—but a new wave of AI wearables has revived interest in bringing these ideas to market.

Enter Deveillance, a startup led by a recent Harvard graduate, with a product called Spectre I that aims to jam always-on microphones in your vicinity. The pitch is simple: if you can’t count on social norms or platform policies to keep your conversations private, carry your own cone of silence.

What happened

Deveillance unveiled Spectre I, a personal device that reportedly emits ultrasonic noise to disrupt microphones on AI wearables around you—think smart glasses, voice-enabled pins, and phones or earbuds that are perpetually listening for activation phrases. The company frames it as a consent technology: you choose when and whether devices nearby can hear intelligible speech.

The promise is enticing, but there’s a catch baked into the physics. Sound doesn’t behave like radio, and modern microphones aren’t naive. Ultrasonic interference decays rapidly over distance, struggles with directionality, and is increasingly filtered or rejected by the audio front-ends in the very gadgets you’d try to defeat. As a result, you can often jam something close to you in a quiet room—but turning that into a dependable, real-world privacy shield is like trying to hold fog in a net.

How an audio jammer is supposed to work

The physics in one paragraph

An ultrasonic jammer floods the air with high-frequency sound (typically 25–50 kHz). Humans don’t hear it, but microphones do—especially commodity MEMS microphones found in phones, glasses, and earbuds. If the ultrasonic level that arrives at the microphone is strong enough, it raises the noise floor or drives non-linearities, making ordinary speech harder to capture or transcribe. Since sound intensity drops fast with distance (inverse square law) and high frequencies are absorbed by air more than midrange speech, the jammer focuses power where it matters most: right next to the mic.

In practice, success hinges on four variables:

• Distance and angle to the target microphone
• Output power and beam shape of the jammer
• Ambient conditions (air absorption rises with frequency and varies with humidity and temperature)
• The target’s audio chain (filters, automatic gain control, beamforming, and machine-learning denoisers)

Modern mics fight back by design

Ten years ago, many consumer microphones happily captured ultrasound and let it alias into the baseband. After high-profile demonstrations of “inaudible command” attacks, vendors tightened things up. Today, many audio front-ends:

• Low-pass or notch-filter above ~20 kHz to reject ultrasound
• Use multi-microphone beamforming to focus on speech coming from the wearer’s mouth and to place nulls to the sides
• Run adaptive noise suppression and echo cancellation trained to keep speech and reject stationary or tonal sources
• Monitor for abnormal high-frequency energy and clamp AGC to avoid overload

That doesn’t make them invincible, but it raises the bar. To win, a jammer must be closer, louder, or more cleverly placed than the target’s algorithms expect—and keep doing that as vendors ship firmware updates.

Why this will be hard to make reliable

1) Range is brutally limited

Ultrasound attenuates much faster than audible speech. On top of the 6 dB loss every time you double the distance (spherical spreading), high frequencies are soaked up by the air, with several decibels of extra loss per meter depending on weather. That’s why almost every successful demo of ultrasonic jamming happens within arm’s length. If the mic you’re trying to blind is on someone’s temple (smart glasses) or at their ear (earbuds), you’d need to be very near—and ideally between the talker and their device’s beamformed lobe—to make a dent. Across a noisy café? Unlikely.

2) Directionality is tricky on a human body

To get decent range without deafening the room, you use an array of ultrasonic transducers to form a beam. But people move, tilt, and turn. Wearables place mics at different heights and angles. A necklace- or lapel-mounted jammer might blast straight ahead while the target mic sits up and to the side; a desk puck could be occluded by a coffee cup. Every small misalignment costs precious decibels where you need them most.

3) Beamforming microphones create off-axis headaches

Glasses, pins, and earbuds increasingly use two or more mics to lock onto the wearer’s mouth and null out side noise. If your jammer sits off-axis, its energy may fall right into those nulls. Even worse, the DSP can adapt, subtracting out stationary interference while keeping transient, mouth-proximate speech. In that setup, jammers often produce a “pumping” effect: they’re clearly present in raw audio but get algorithmically sidelined while the target’s speech sails through.

4) Collateral damage—and not just to gadgets

An ultrasonic fog bank doesn’t honor your intent. It may:

• Disrupt your own devices’ voice control and dictation
• Trigger or degrade assistive devices with microphones, such as hearing aids or personal amplifiers
• Disturb pets (many animals hear well into ultrasonic bands)
• Generate audible artifacts via intermodulation, leading to ear fatigue or headaches for some people

What begins as a privacy tool can become, practically speaking, a public nuisance—especially in shared spaces.

5) The cat-and-mouse game favors platforms

As soon as users start relying on jammers, device makers can ship countermeasures: steeper front-end filtering, ultrasonic classifiers that auto-mute interference, or even cooperative beacons that let their devices ignore known jammers. Platforms iterate via software updates; hardware startups must ship new units. That asymmetry is harsh.

6) Power, heat, and comfort are nontrivial

Projecting a meaningful ultrasonic field requires power. More power demands larger batteries and better thermal management. A wearable that runs warm, hisses around pets, and needs a midday recharge is a tough sell—especially compared to simply long-pressing a mute button or using a social cue.

What it means for policy and etiquette

The impulse behind Spectre I is valid: the burden of privacy should not rest only on bystanders’ politeness or the goodwill of platform companies. But the limits of ultrasound point to a bigger truth: if society wants conversational privacy in a world of ambient AI, we will need layers beyond gadgets.

• Product design: Clear opt-in indicators, unmistakable recording lights, and hardware kill switches change norms more effectively than arms races.
• OS-level policy: System-wide “no capture zones” and APIs that apps must honor can put real teeth behind consent signals.
• Standards: A voluntary “Do Not Listen” broadcast (via Bluetooth or ultrasonic watermark in the audible band) would be far more effective if big platforms commit to honoring it. Without that social contract, any beacon is wishful thinking.
• Law: Existing wiretapping and eavesdropping laws predate wearables. Legislatures can clarify when ambient capture is permissible, what notice is required, and how bystanders can seek redress.
• Etiquette: Just as we learned not to put calls on speakerphone in crowded trains, we can learn to announce recording, remove always-on glasses indoors, and respect posted “no audio” zones.

Jammers can be provocative art, research tools, and conversation starters. But as primary protection, they’re more warning flare than shield.

Key takeaways

• Ultrasonic jammers like Spectre I aim to raise the noise floor at nearby microphones so speech becomes unintelligible to AI assistants.
• Physics is the main adversary: ultrasound decays quickly with distance and is strongly absorbed by air, limiting useful range to close quarters.
• Modern wearables use filtering, beamforming, and ML denoising that blunt simplistic jamming, especially when the jammer is off-axis.
• Collateral effects include disrupting your own devices, bothering pets, and creating social friction—complicating everyday use.
• Even partial success will likely prompt platform countermeasures delivered via software updates, resetting the game in their favor.
• Sustainable privacy against ambient AI will rely on design norms, platform policies, legal frameworks, and social etiquette—not just gadgets.

What to watch next

• Firmware updates on major wearables: Do vendors harden against ultrasound more aggressively as jammers hit the market?
• Standardization attempts: Any cross-industry moves toward a “consent beacon” for audio (BLE, Wi‑Fi Aware, or audible watermark) that devices agree to respect.
• Regulatory signals: Guidance from agencies on acoustic interference in public places, and updates to consent and eavesdropping statutes.
• Assistive tech impacts: Research clarifying how ultrasonic fields affect hearing aids and cochlear implants in the wild.
• Product evolution: Second-generation jammers that use smarter beam steering, situational awareness (ultrasonic LIDAR-style mapping), or hybrid approaches (audible masking plus ultrasound) to improve reliability without undue nuisance.
• Social norms: Whether venues, workplaces, and transit systems establish clear audio-capture policies—and whether they’re enforced.

FAQ

Will an ultrasonic jammer stop smart glasses from recording me?

Sometimes, at short range and favorable angles—but not reliably. Glasses typically point microphones toward the wearer’s mouth and use DSP to reject side noise. Unless you’re very close and your jammer is aimed well, speech may still get through.

Is using an ultrasonic jammer legal?

In many places, emitting ultrasound isn’t specifically regulated like radio jamming. However, you can still run afoul of noise ordinances, workplace rules, or civil nuisance claims. Intentional interference with others’ assistive devices could also raise liability issues. Check your local laws.

Can animals hear these devices?

Yes. Dogs, cats, and many other animals hear into the ultrasonic range. High output levels can bother or stress them. Ethical products should include safety limits and clear guidance on use around pets.

Will it jam hearing aids or cochlear implants?

Modern assistive devices typically focus on the audible band and may filter ultrasound, but strong ultrasonic fields can still affect automatic gain control or cause artifacts. Real-world impact depends on the device and environment; more independent testing is needed.

Why not just filter ultrasound on wearables and call it a day?

Many already do, but the world is messy. Ultrasonic energy can mix down via non-linearities, and some microphones still capture beyond 20 kHz. More aggressive filtering can also degrade desired audio quality. It’s a balancing act.

Are there better alternatives to protect my conversations?

• Use device-level privacy features: hardware mutes, explicit consent modes, and visible indicators.
• Favor products with clear privacy design (e.g., recording lights that can’t be disabled).
• Adopt social cues and policies: ask before recording, remove always-on devices in sensitive spaces, and respect posted rules.
• For facilities, consider acoustic design (sound masking in specific rooms) that is audible but gentle and well-signposted.

Could a jammer broadcast a “Do Not Listen” beacon instead of noise?

Technically yes—via Bluetooth, Wi‑Fi Aware, or even an audible watermark—but it only works if platforms honor it. Without vendor buy-in, it’s a polite suggestion devices can ignore.

Is ultrasound safe?

The consensus for low to moderate levels is generally yes, but very high levels near the body raise comfort and safety concerns. Standards exist for occupational exposure, and consumer devices should stay comfortably below them. If you feel pressure, headaches, or discomfort, stop using it.

Source & original reading

Original article: https://www.wired.com/story/deveillance-spectre-i/