End-to-end encryption is the most obvious misconception. It makes people think there’s no way for bad actors to see what’s going on in the middle, but in order to secure a video call, something in the middle has to be encrypted too, not just at each end. It’s a threat model you have to figure out for yourself. If you’re a book club, you probably don’t need security, but if you’re having a board meeting, you probably do. You’re likely fine 99% of the time but it’s that 1% that matters.
Everyone thinks encryption is easy and they’re right – it’s super easy to encrypt something. What’s hard is verifying who encrypted it. We’ve tried to simplify this by showing a visual indicator so that a user can confirm that the communication wasn’t intercepted. By matching a visual password embedded into the video stream with one that the user knows, they’re able to confirm that the session is encrypted without having to trust us, the video conferencing provider.15 November 2020
From background noise suppression and echo cancellation to voice and facial recognition, machine learning has already made huge strides in simplifying the way we communicate from an end-user perspective. As machine learning programs become increasingly sophisticated, these use cases and opportunities will continue to grow at an incredible rate.12 December 2019