Walk your trading floor or your ops bullpen this week and look at people’s faces. If anyone is wearing camera-equipped glasses at a workstation, and they increasingly look exactly like ordinary Ray-Bans, you may have a problem your cybersecurity policy was never written to catch.
Meta’s camera-equipped glasses, the Ray-Ban Wayfarer line, the newer Ray-Ban Display, and the Oakley Meta models, have crossed from novelty into mainstream. They look like ordinary eyewear. They record photo, video, and audio. And they connect to a consumer cloud environment entirely outside your firm’s contractual control. Reporting this year established that content routed through the glasses’ AI features can end up in front of human reviewers overseas, and Meta’s voice-recording retention terms have changed repeatedly and are not something you get to negotiate.
But the vendor’s data practices are actually the second problem. The first problem is simpler: think about everything a camera at eye level sees during a workday.
The capture surface is your entire firm
Every security control you’ve built assumes data leaves through a channel you can watch: email, endpoints, USB ports, print queues, network traffic. A camera on someone’s face bypasses all of it. Walk through what’s in frame:
Screens. Client account numbers. Positions and holdings. Trade blotters. Wire instructions with routing and account numbers. Custodian portals. Portfolio models. Fee schedules. Email threads with counsel. Your CRM, open to a client’s entire financial life. Whatever is on the monitor is in the recording, at whatever resolution the current generation of these cameras captures, which is more than enough to read a screen at normal working distance.
Credentials. A password typed while the glasses watch the keyboard. An MFA code displayed on a phone. A one-time passcode read off a token. A password manager unlocked with the vault visible. Screen contents are bad; captured credentials are worse, because they don’t just leak data; they hand over the keys.
Phones. Everything above applies to the mobile device in the employee’s other hand. Texts from clients. Authenticator apps. Banking apps. The phone screen is smaller but it’s held closer.
Conversations. These devices record audio, and audio doesn’t need line of sight. The portfolio manager discussing a pending trade two desks away. A client call on speakerphone. The partners talking through a difficult client situation in a hallway. An investment committee meeting where someone in the room, or someone who walked past the open door, is wearing them. For firms handling MNPI, every sensitive conversation within microphone range is now potentially a recording sitting on a personal device connected to a consumer cloud.
Everything else at eye level. Whiteboards after a strategy session. Printed statements on a desk. The visitor log at reception. Sticky notes that shouldn’t exist but do. Badge numbers. The office layout itself, which, if you think like an adversary for a moment, is reconnaissance.
And there are two capture paths, not one. The accidental path: an employee wears them all day because they’re their prescription glasses now, and sensitive material simply accumulates in their camera roll and voice history. The deliberate path: these are a near-perfect exfiltration tool. No USB port to block, no DLP rule to trip, no print log, no email to flag. An employee who wants to walk out with your model portfolio or client list just has to look at the screen. Your entire monitoring stack is blind to a camera.
“But there’s a recording light”
Who cares? Be honest about what that LED actually is: a small light, on a device the wearer owns, that the wearer controls.
Everyone owns black electrical tape. Beyond the low-tech option, there’s a market of third parties who will disable the indicator outright, visible enough that California introduced a bill this year (SB 1130, pending) targeting exactly that. And even if the light works perfectly and nobody covers it: who at your firm is assigned to monitor whether a colleague’s glasses are glowing? Nobody. That’s not a control, that’s a decoration. No examiner will accept “the device has a light” as a safeguard, and neither should you.
The indicator also does nothing for audio. A voice interaction or ambient recording involves no visible signal your staff could catch even if they were looking.
Where this lands with the regulators — briefly
You don’t need a treatise here, just the mapping:
- Reg S-P (as amended 2024): an unmonitored camera capturing customer information and transmitting it to an unvetted consumer cloud is precisely the kind of foreseeable risk your safeguards and incident-response program is supposed to address.
- Rule 206(4)-7: if this risk isn’t addressed anywhere in your policies, personal devices, recording, confidential-information handling, that’s a gap in a program that’s supposed to be reasonably designed.
- Section 204A / your code of ethics: captured audio and screens containing MNPI belong in your insider-trading and information-barrier controls.
- Rule 204-2: if the glasses are used to create or send business communications, you’ve got an off-channel records problem on top of everything else.
That’s the whole regulatory story. The interesting part is above it: the capture surface.
The objections you’ll hear — and why they fail
Take this to your Managing Partner or your IT committee and you’ll get pushback. Here’s the pushback, steel-manned, and here’s why the prohibition survives it.
“Everyone already carries a phone with a better camera. We don’t ban those.” This is the best counterargument, so take it seriously. The answer is detectability. Photographing a screen with a phone requires a deliberate, socially visible act: pull it out, raise it, aim it. A colleague sees it; a manager can interrupt it; a camera in the ops area catches it. The glasses capture whatever the wearer’s face is pointed at, continuously, with zero observable behavior. The act of looking is the capture. Your entire control environment, supervision, clean-desk culture, the basic social friction of pointing a camera at something, is built on capture being a visible act. These devices remove that assumption. Same sensor as a phone; completely different threat, because the phone’s risk is governed by behavior you can see and the glasses’ risk is not. (The related objection: “a determined insider could memorize or photograph things anyway” — proves too much: by that logic you’d unblock your USB ports too. Controls exist to raise the cost and observability of exfiltration, not to make it metaphysically impossible.)
“Nobody has ever been cited for this. Where’s the incident?” Two answers. First, that was equally true of personal texting in 2019, and the firms that waited for the first enforcement action ended up funding the off-channel sweep. Regulators reward the firms that addressed a foreseeable risk before it had a case name. Second, and more uncomfortable, an exfiltration channel that’s invisible to every monitoring tool you own produces exactly zero detected incidents whether or not it’s being used. The absence of evidence here is what the blind spot looks like from the inside.
“You can’t enforce it. Prescription smart glasses look like regular glasses.” True, and irrelevant. Most of your compliance framework already runs on attestation plus consequences rather than perfect detection: personal trading, outside business activities, gifts and entertainment. Nobody argues the code of ethics is pointless because you can’t surveil every brokerage account in real time. The policy does three things detection can’t: it removes the “nobody told me these counted” defense, it creates a clean disciplinary and termination basis when a violation surfaces, and it documents for the examiner that the firm evaluated the risk. A written prohibition your staff has acknowledged is a control. A light on someone’s face is not.
If your firm hears an objection not on this list, it’s probably a version of “the vendor’s cloud is secure.” That one misses the point entirely: the issue isn’t whether Meta’s data center is well-defended, it’s that customer information reached a third party with no agreement, no due-diligence file, and no deletion rights. Reg S-P doesn’t grade the unauthorized recipient’s security posture.
The policy answer
This is one of the rare cases where nuance is the wrong instrument. Don’t write a tiered, permission-based framework. Write a prohibition:
No employee, contractor, or visitor may wear or operate camera-equipped eyewear or other wearable devices capable of capturing photographs, video, environmental audio, or screen content while accessing firm systems, viewing or discussing customer information or material nonpublic information, handling authentication credentials, or participating in confidential firm meetings. This applies in the office, at remote work locations, and anywhere firm work involving such information is performed. Devices must be removed from the immediate work area, not merely powered down while worn, before protected information is accessed, viewed, or discussed. The firm will evaluate reasonable-accommodation requests and legitimate business uses through Compliance, HR, and IT on a documented, case-by-case basis.
Why “removed from the work area” instead of “powered off”? Because you can’t verify the power state of an opaque consumer device, the same reason the LED is worthless. “The glasses were on the shelf across the room” is a control someone can actually attest to. And the accommodation clause isn’t politeness: assistive hearing devices with microphones exist, and a blanket ban with no exception process trades a security problem for an employment-law one.
Making it stick:
- Amend your Employee Technology Usage Agreement and safeguards policy now, not at the annual review. Date the amendment. You want a 2026 timestamp when an examiner asks.
- Cover remote work explicitly. The glasses at the kitchen table with a VDI session open are the same exposure, arguably worse, since nobody’s watching.
- Give visitors operational controls. Reception signage, a visitor acknowledgment, host responsibility, storage before entering ops areas, and a clause in your vendor MSAs. The MSP tech doing a site visit in Meta Oakleys is inside your perimeter with a camera.
- Train on it. People genuinely do not think of these as recording devices; they think of them as sunglasses. One FieldCraft-style module or a five-minute segment in annual training closes that gap.
- Document enforcement. A policy you identified as necessary and then didn’t enforce is a self-authored deficiency finding.
The bottom line
Everything your firm protects, screens, credentials, conversations, documents, passes in front of employees’ eyes and ears all day. These devices put a camera and microphone in exactly that spot, feeding a consumer cloud you don’t control, behind an indicator light defeated by a strip of electrical tape. The capture surface is total and your monitoring stack can’t see any of it.
The control is about as cheap as controls get: a paragraph in your P&P manual, a paragraph in your tech usage agreement, some signage, and five minutes of training. The alternative is explaining to the SEC, or to a client, how their account statement ended up in a data center you’d never heard of.
If you want a second set of eyes on how your current safeguards policy handles body-worn devices, that’s exactly the kind of gap an MTradecraft risk assessment is built to catch.
BrainTrust members: the same physical-device exposure runs through The Unseen Risk of a Stolen iPhone to Your Firm’s Network, and the safeguards obligations behind this policy trace back to The New SEC Regulation S-P Amendments: What Every RIA Needs to Know. The policy templates in the library are where the prohibition above belongs.
MTradecraft LLC is a vendor-independent cybersecurity compliance firm in Dallas, TX serving SEC-registered investment advisers, hedge funds, broker-dealers, and family offices. Nothing here is legal advice.