The device points both ways
A pair of camera glasses is usually described by what it collects about the world in front of it. It also collects a continuous record of where its owner went, what they looked at long enough to ask about, who they were near, and what they wanted to know at each moment. That record is more intimate than anything the camera catches of a stranger, because it is not a fragment. It is a sequence.
EFF has been documenting the state of consumer wearables generally, including a review this summer finding that most smart watches, rings and bands lack basic transparency reporting and key privacy features. The instinct behind that work applies with more force to a device carrying a camera, a microphone and an assistant.
Two arguments that are usually kept apart
The industry has an interest in treating wearer privacy and bystander privacy as a trade-off, where protecting one costs the other. Notice for strangers gets framed as friction for owners, and owner control gets framed as an obstacle to safeguards.
They are not opposed. On-device processing serves both. Short retention serves both. A published account of what leaves the device serves both. The only party that loses when those become defaults is the one collecting, which is why the two constituencies keep being told they are in conflict. A wearer asking where their own footage goes and a bystander asking the same question want the identical answer.
What our own tool cannot do
We build a detection app, so we should be plain about its limits here. It tells you something about devices near you. It tells you nothing about what the device on your own face is doing, and it cannot, because that information never appears on the air.
If you want to know what your glasses send home, the only sources are the manufacturer's documentation and whatever independent testing exists, and there is not enough of the second kind. That is a gap somebody should fill, and it is not a gap a scanner can close.