How Private Are Your AI Glasses, Really?

Key takeaways

• AI glasses privacy relates to how your personal data is processed by AI providers, and, if they have a camera, making it clear when you’re recording and abiding by all local laws
• Privacy-centric AI glasses, such as the HTC VIVE Eagle, process and keep your data on device and smartphone, and on a private cloud servers when necessary, rather than sharing your prompts or media for training their AI models
• On-device processing means prompts spoken or media taken with AI smart glasses are kept offline and only shared between the glasses and the smartphone whenever possible, unless it’s a complex query
• Not using your prompts to train models means, by default, your interactions with AI aren’t passed to any data servers for storage. If your question requires online queries, those prompts should be stripped of personally identifiable information by the intermediary so it’s harder for an LLM to build a profile of you
• It’s legal to wear smart camera glasses in public if all local laws are followed regarding data capture. Etiquette matters though: CNIL in France found 67% of people surveyed ‘consider smart glasses to pose a potential privacy risk’. So it’s important to always check regulations before any recording, and avoid sensitive situations
• The most privacy conscious AI smart glasses have a bright, clear LED that shows when you’re recording or taking photos

What does privacy in AI glasses mean?

Privacy in AI glasses covers two main areas, data protection and social transparency: 

Whether your data and media are kept on device or processed in the cloud. 

You should be aware whether your prompts are used to train AI models, are looked at by a human, and where your inputs are kept, to truly be in control of your data.

HTC VIVE Eagle is designed with this in mind: your data isn’t used for training the VIVE AI model or any models it uses, and more complex prompts sent to third-party AI providers are de-identified so digital profiles are hard to build. Its use of third party models are via enterprise services, which do not train off the queries. 

Respecting the privacy of others when wearing AI glasses. As smart glasses with cameras are growing in popularity, some people are worried about being recorded without consent, where at least two thirds of people would move away when smart glasses are used to take video or pictures in sensitive situations. 

HTC has built protection into VIVE Eagle to make it clear when photos or video are being taken, with the prominent LED status light pulsing when recording, flashing once when photos are taken, and giving the subject that ability to stop the recording if they wish. 

How is my data processed?

AI prompts on smart glasses are processed with differing degrees of privacy. Using photos and videos as an example, your data is either:

• Held entirely on device and shielded from cloud storage or human review (such as on HTC VIVE Eagle)
• Uploaded for cloud-based processing, such as Meta AI, when using the service on compatible AI glasses

A study by Pew Research found 67% of respondents say they “understand little to nothing about what companies are doing with their personal data”, so it should be simple for you to understand what’s happening with your sensitive information.

HTC takes a ‘privacy-first’ approach to its glasses, giving clear information about how data is collected, processed, stored and shared.

Your photos and videos aren’t shared with the cloud - instead, they’re kept safely on your AI glasses or smartphone.

The same is true of commands made to the on-board VIVE AI, where your voice data is kept safe and offline.  If you want to ask more complex queries, your queries will be diverted to use Google’s Gemini and OpenAI’s GPT enterprise models.

Some AI-enabled features may require data to be transmitted to supported services for processing. However, your requests to these services are de-identified to make it more difficult for LLMs to build a profile of you. This makes it harder to link queries to your personal data.

Here’s a table that shows the difference between AI glasses that are privacy-first and put the control of data in the hands of the user, such as VIVE Eagle, and those that take a cloud-based approach, with data available to train AI models.

How does data privacy compare between VIVE Eagle and other smart glasses, like Meta AI smart glasses?

Voice prompt sharing - VIVE Eagle - Stored locally or de-identified before user chooses to share information with third-party LLMs. Real-time translation and meeting transcriptions are sent directly to the cloud for faster processing, accuracy and increased language support. 

Voice prompt sharing - Meta AI glasses - Voice transcript shared to cloud processing to improve AI performance, when using Meta AI functions and voice recordings and text transcripts storage is enabled. 

Media processing - VIVE Eagle - Held in 32GB on-device storage or transferred by user to smartphone. No training for LLMs. 

Media processing - Meta AI glasses - Processed via the cloud when uploaded to the app, stored for 30 days. No training for LLMs. 

Human review - VIVE Eagle - User data or media is never used for training purposes or reviewed by a human being. 

Human review - Meta AI glasses - Humans are occasionally used to verify and label voice recordings and text transcripts when opted in. 

Supported AI systems - VIVE Eagle - Google Gemini, OpenAI GPT, VIVE AI 

Supported AI systems - Meta AI glasses - Meta AI

Voice data storage - VIVE Eagle - On device whenever possible. Real-time translation and meeting transcriptions are sent directly to the cloud for faster processing, accuracy and increased language support. 

Voice data storage - Meta AI glasses - On Meta cloud servers for one year when opted in 

Encrypting data on AI glasses 

The most privacy-conscious AI smart glasses use high-level cybersecurity - such as the AES-256 used with HTC VIVE Eagle . This is advanced, exceptionally strong encryption - in fact, so strong that it’s recommended by the US government as a cybersecurity standard.

This level of hacker protection means that should you lose the smart glasses, then nefarious actors won’t be able to get into the device and gain any access to your information or media.

Without authentication from your VIVE Connect app, your data is locked away from any thieves and remains secure on the AI glasses.

The most private AI smart glasses combine these high levels of encryption with on-device processing to ensure the wearers know that their data is secure. 

HTC ensures that any personal data remains in your control - in the event you choose to use a third party AI tool for more complex tasks, such as Google Gemini or OpenAI GPT, your data is scrubbed for personally identifiable information before being sent over.

Why the camera LED is so important for AI glasses privacy

Smart AI glasses should have a front-facing, bright LED that informs users of video or photos being captured, so bystanders can choose whether to appear on camera.

HTC believes strongly in protecting privacy, which is why the VIVE Eagle has been designed as ‘privacy-first’ AI glasses.

The HTC VIVE Eagle AI glasses enhance camera privacy by using a dual-LED approach - a pulsing, bright, forward-facing LED to notify bystanders of camera use, and an internal LED that alerts the wearer that the camera is recording.

The forward-facing LED state is key to privacy: when the camera is operating, a light will always illuminate to show that a picture is being taken (a single flash) or a video is being recorded (steady pulsing light).

The internal status light allows the user to see instantly when images or video are being taken, to alert them should any accidental footage be captured.

If anyone attempts to cover this light, or disable it by tampering with the AI glasses, the camera will no longer function. The glasses must always be worn when recording - if the user attempts to set them down when capturing video or photos, the device will disable the camera.

Is it legal to wear camera glasses in public?

It is generally legal to wear AI glasses with cameras in public, but you should always familiarise yourself with the relevant laws and requirements in your location, which are constantly changing and evolving.

Even if you’re complying with local regulations, wearing AI camera glasses responsibly is about etiquette - and be respectful about takings photos and videos which feature other people. A study from China showed that 80% of people in a mall don't want to be recorded through smart glasses. 

Consider these good practices when using camera-enabled AI glasses in public:

• Remove your AI glasses in sensitive areas, such as medical or religious settings, changing rooms, or when working with children
• Ensure you don’t record video or take photos that could reveal personal information or where you could accidentally infringe on a person’s privacy
• Always ask before capturing images or video of someone in public, unless you have explicitly obtained permission
• Make it clear when you’re about to begin capturing media to ensure those around you have the choice whether to appear on camera

You can check guidance like that from CNIL (the national data protection authority for France) on ways to protect the privacy of those around you when wearing smart glasses, and see if your country has similar guidance.

The same applies to audio recording - check the local laws governing recording conversations, but for the avoidance of doubt you should always ensure you have the consent of everyone speaking before recording to ensure you’re adhering to the law and complying with workplace, venue or other policies that might be in force.

How to choose AI glasses that put privacy first

When buying AI glasses that are private with your (and other’s) data, use this checklist to help make your decision: 

• Do the AI glasses have physical hardware locking? This means if your glasses are stolen, your data cannot be seen by anyone else
• Does the camera work if the recording LED is covered? The AI glasses should shut off power to the camera if the LED is covered, tampered with or the glasses are removed
• How secure is the data on your AI glasses? Look for military-grade encryption, such as AES-256: it should be virtually impossible for hackers to bypass the hardware locks and steal your data
• Can my prompts be used to train AI models? Look at the smart AI glasses’ data training guardrails - your data (such as voice prompts, media or other personally identifiable information) should never be sent to train AI data models, and any identifying data should automatically be scrubbed when you share it with third party LLMs
• Will a human look at my images or hear my voice prompts? Privacy in smart glasses should also extend to human review - other people should not be able to view your media or prompts, even to improve services, without your explicit permission