This would be great as long as it's on device only.

This post raises an important point about how useful technology doesn’t have to come at the cost of personal privacy. The desire for on-device facial recognition mirrors broader discussions about responsible innovation—using advanced tools in a way that respects user boundaries. That balanced mindset is something often reflected in analytical spaces like the Cliff Mass Blog, where complex systems are discussed with transparency and care. In a similar spirit, conversations around privacy-friendly AI deserve thoughtful attention, much like the science-based discussions found at https://cliffmass.blog/ .

The emphasis on local processing instead of cloud-based solutions is especially relevant today, as users become more aware of how personal data is handled. Privacy-first AI design is gaining traction across multiple fields, and it’s encouraging to see these ideas applied to everyday tools like photo organization. For readers interested in ethical AI and privacy-respecting technology development, the Mozilla Foundation offers useful insights and research at https://foundation.mozilla.org , which aligns well with the concerns raised in this post.

If offered it should absolutely be optional, not default, and should be OFF by default. One thing I feel very strongly about is ensuring that all privacy-concerning features should be off by default, and require an action on the part of the owner to turn them on, and if then turned off again later ALL associated data should go poof and be stored nowhere at all.

I know this is possible as ente.io does it, alongside scene labels (ie. mountains in the evening). I tried their product a while back and it wasn't bad at all. Probably would take a lot of work and time but would be super useful.

Would love something like this, but I would suggest: the labelling should be explicit. The user should be able to add or remove any the labels that the AI adds automatically.

I hardly can find photos now in the photos tab, so as a workaround i save my photos in Proton drive now in a folder structure so i can find them later on. So this feature would really be great for me too.

This would be a very nice feature. I would add not only face recognition, but general object / situation awareness, to make search on par with what other services like Google Photos offer (for instance, searching "car" brings up every photo with a car in my album).

I guess it is a challenge to implement in a privacy-respecting way, as the entire library would probably have to be processed on device and then you would either have to trust Proton to store the index, or reprocess everything again in each device.

Essential. Really need facial recognition or I wouldn’t be able to find anything! Got to be on device I guess, so understand it’s non-trivial. (Although I’d sacrifice e2e encryption of photos for this myself.)

Yes. Absolutely necessary.

I disagree with this feature. I wouldn't want to see more AI baked into my services and products. Thanks.

This is so important!! Ente Photos have a machine learning process running locally for this reason. I really want a similar experience with Proton. However, judging from the slow approach to linux, I don't think we will have this feature anytime soon :(

I would like to see an option to toggle (for privacy/security it should be off by default) that recognizes recurring people in images, and allows you to add their name to their face so that they are automatically recognized in the future. Perhaps this could use some AI model, or machine learning technique. It would be interesting to see this implemented in a secure way, say the AI model could be self hosted, or hosted on a proton server with zero-access encryption.