Meta is reportedly working on a pair of smart glasses that would take photos and record audio every few seconds, enabling its AI service to become more aware of its surroundings and help the wearer recall things.
In tests of the prototype, the red LED light on the corner of the glasses does not turn on, meaning non-wearers would not know when the glasses are recording. This has reportedly caused internal friction at Meta due to the obvious privacy concerns.
Meta CEO Mark Zuckerberg has said he wants the company’s smart glasses to act more like an agent for the wearer, rather than just a pair of glasses that can take photos and videos. Adding more ambient awareness could make the glasses more useful, increase usage time, and potentially turn them into a more credible alternative to the smartphone.
Getting around the privacy concerns
According to reporting by the Financial Times, Meta has been working on several ways to reduce these privacy concerns. One system would turn footage and audio into anonymized metadata before it reaches Meta’s servers, ensuring that neither Meta nor the wearer can access the raw footage. Whether that would satisfy data privacy and wiretap laws will ultimately be up to regulators.
Meta does not have the strongest reputation for user privacy. It has faced several privacy and addiction lawsuits over the past few years, and recently had to halt an employee tracking program after sensitive data was exposed.
Meta’s current smart glasses do not flash when the wearer asks AI questions that require camera access, although there is still an indication via the wearer’s verbal prompt. This next step would go much further, with images and audio captured without any prompt or interaction.
Meta invests heavily in smart glasses and AI
Meta has put significant resources behind smart glasses, which reportedly sold 7 million units last year. It has also launched cheaper versions of the product ahead of expected competition from Apple, Google, and Samsung.
At the same time, Meta has invested billions of dollars in AI data center infrastructure. It raised its 2026 capital expenditure guidance to $125 billion to $145 billion and has made major investments in neoclouds to secure more compute capacity.
That has helped Meta earn decent reviews for its latest AI model, Muse Spark. However, actual interest and usage of Meta’s AI apps appear to be behind the leaders in the field, including OpenAI, Anthropic and Google. This has reportedly led Meta to explore selling excess compute capacity to rivals, which could help offset part of the cost of its AI expansion.
If Meta manages to get this smart glass prototype to market, it could have far-reaching privacy implications for all app developers building apps for smart glasses. Developers will need to take extra care to ensure faces are blurred, audio isn't captured on servers, and that Meta's push for agentic service doesn't damage their own brand.
Read more: Meta is also exploring ways to monetize its massive AI infrastructure through a potential cloud compute business. Here's what that could mean for enterprises.