Ethics & Privacy
Video of you moving is personal data. Here is what we do with it, in plain language, and where the formal policy lives.
Mocta-X Dashboard processes video of you moving. That's one of the most personal data streams a piece of software can touch, and we take that responsibility seriously — not as a line item in a privacy policy, but as the shape of the architecture itself.
The honest answer to "where does my video go?" is: to our servers, to be analysed, and nowhere else. It is stored privately so you and the coaches you choose can replay the analysis, it is never sent to an AI provider, and it goes when you delete it. The full commitments are in the privacy policy (v2.0, effective September 15, 2026); this page is the plain-language version of the ones that matter most.
Our four commitments
We say exactly where the video goes
Uploaded clips go to our servers to be analysed, are stored privately behind time-limited links, and are deleted with your account. Realtime practice on the web never uploads video. The iPhone app can run pose extraction on-device (Settings → Pose model) so only joint positions leave the phone.
You own your sessions
Export your data, correct it, or delete your account and everything with it from Settings. Deletion begins immediately; we work to complete removal from live systems within 30 days, after which backups age out.
Encrypted, and never sent raw to an AI
Data is encrypted in transit and at rest. When the AI coach answers, it receives the metrics, scores and history it needs — never your raw video or images — under terms that forbid the provider from training on it.
AI that shows its work
The coach investigates the graded data before it writes and shows the trail. Every claim names the stroke and frame it came from, and every metric carries the confidence the capture supports.
How that shows up in product decisions
Commitments matter when they cost something. These are three places we chose the harder path because it was the right one.
No face recognition. No biometric template.
The pipeline identifies joints, not people. We do not build a biometric template of you, and we do not use motion data to identify you or look for you in anyone else's footage. Gait can in principle be identifying, which is exactly why any de-identified data used to improve the models is separated from identity first and never re-linked.
What never trains a model
Raw video and images, any frame of your face or body, your profile, injury notes, goals, chat messages, coach comments, and anything from an athlete a club has identified as under 16. We do not sell, license or share training data, and you can opt out of model improvement entirely.
We publish our limitations
The methodology page lists where single-camera pose estimation breaks — occlusion, depth, frame rate, non-determinism — and what the product does about each. If a coach asks 'should I trust this score?', the honest answer is one click away, and the app withholds a grade it cannot support.