Stryv Responsible AI System — Active

AI Built To Serve
Human Effort.

STRYV uses artificial intelligence to understand movement, protect competition and create better fitness experiences — without taking ownership away from the athlete.

Your body performs.
Our AI verifies.
You earn the result.

// The Trust Mark

One Mark.
One Standard.

What this mark means

This mark identifies STRYV features that meet our standards for human control, privacy, fairness, security and responsible AI use.

Position

The Athlete Is Not The Data.
The Athlete Is The Purpose.

STRYV does not use AI because it is fashionable. We use it where it can make physical competition more accurate, accessible and engaging. Every system must justify its place in the experience.

AI should verify —
not invent.

AI should assist —
not override.

AI should explain —
not hide.

AI should protect —
not exploit.

AI should serve people —
not reduce them to profiles.

Architecture

Human At The Centre.
AI At The Edge.

Three layers, one hierarchy. The athlete holds the centre. The AI layer detects, measures and verifies around them. The governance layer watches the AI. Select any layer to see its job.

The Athlete Movement · Consent · Control
AI Layer

Detect

AI identifies movement patterns required to support rep tracking and exercise analysis.

Wording is aligned with the production system: movement analysis runs on-device and raw video is never stored, in the browser or the app.

Systems

AI With
A Job To Do.

Every AI system inside STRYV has a defined purpose and an honest status. LIVE means shipped. Everything else is labelled exactly as it is.

The Standard

The STRYV
Responsible AI Standard.

Eight requirements for any AI-powered STRYV feature. As each one locks in, the approval badge moves closer to its verified state.

0 / 8 Standards Verified

Camera Privacy

Your Camera Is A Sensor —
Not An Audience.

STRYV uses camera input to understand physical movement during supported experiences. Our goal is to process only what is needed to deliver the feature, while giving players clear information and meaningful control.

Nothing is recorded. Nothing is uploaded. Nothing is kept.
Raw video stored: 0 bytes.

Browser 1v1 battles Verified in code — pose detection runs inside your browser on your device. The battle page contains no code that uploads video or camera frames anywhere.
Pose model delivery Verified in code — the browser battle loads a standard pose-landmark model to your device; the model provider serves the model file, it does not receive your video.
Mobile app processing Confirmed — movement analysis runs on your device in the app, the same way it does in the browser. Camera frames are processed in real time and never leave your phone.
Raw video storage Confirmed — raw video is never stored. Frames are analysed live and immediately discarded. Nothing is recorded, saved or uploaded.
Retention period Confirmed — zero. No camera footage or movement video is retained. Only your results — like rep counts and rank — are kept with your account.
Model training policy Confirmed — your camera footage is never used to train models. It is never stored, so it cannot be.
Third-party processors Confirmed — no third party receives your camera data. The pose model runs locally on your device.
Security

Security Is Part
Of The Product.

No single control is treated as sufficient. STRYV's security approach combines product design, technical controls, restricted access, monitoring and continuous review. Expand any layer for detail.

Fair Play

One Standard.
Every Player.

A ranked fitness platform only works when achievement is earned. AI-assisted verification should support consistency while recognising that bodies, environments, cameras and devices can differ.

Body Proportions Camera Angles Lighting Conditions Device Quality Clothing Movement Styles Accessibility Needs Connectivity Valid But Unusual Technique

No model is flawless. Our responsibility is to test, measure, improve and provide routes for review when the system gets something wrong.

Transparency

No Mystery
Decisions.

When AI affects your result, you should be able to see why — in plain language — and challenge it. This is how a reviewed decision should look.

Rep Not Counted
Reason Full extension was not detected
Expected Arms reach full extension at the top of the rep
Detected Partial extension before the next descent began
Confidence 82%
Try Move the camera slightly lower and keep your full body visible
Honesty

Intelligence
Requires Honesty.

AI can be powerful without being perfect. Responsible systems clearly communicate where they can fail — so here is where ours can.

Movement can be misunderstood — especially fast, partial or overlapping motion.

Unusual camera angles can reduce accuracy. Side-on framing works best for push-ups.

Poor lighting can affect detection. The system may miss reps it cannot clearly see.

Temporary technical issues can occur — devices, browsers and networks all vary.

Some bodies and movement styles may be underrepresented in pose models. We treat this as a testing priority, not a footnote.

Confidence varies rep to rep. Updates can change system behaviour, and human review may sometimes be required.

Research

STRYV
Frontier Lab.

Where we explore what's next. Nothing here is a released feature — statuses are exact.

Research does not automatically become a product feature. New systems must demonstrate a clear benefit and meet STRYV's Responsible AI Standard before release.

Accountability

Who Watches
The System?

Responsible AI requires active ownership. Every AI-powered system should pass through human checkpoints — before release and for as long as it runs.

01Assigned Owner A named person is responsible for each AI system.
02Documented Purpose What the system does, and why it deserves to exist.
03Pre-Release Testing Reliability and limitation testing before players ever see it.
04Security Review Controls checked against the security architecture above.
05Privacy Review Data use checked against purpose limitation and minimisation.
06Performance Monitoring Live behaviour watched for drift, failures and unintended outcomes.
07User Feedback Player reports feed directly into review and retesting.
08Incident Escalation Problems have a defined path to a human who can act.
09Periodic Reassessment Systems are re-reviewed — and retired if they no longer meet the standard.
Human Approval Required

Automated systems are governed by people. That is the final checkpoint, always.

Changelog

AI Transparency
Log.

Meaningful AI-related product updates, recorded in the open. Entries below are sample placeholders until real updates are published.

Trust Centre

Access
Points.

Disclosure

Found Something We
Should Know About?

We welcome responsible reports about security weaknesses, privacy concerns, unfair AI behaviour or unexpected system results. Reports go directly to the STRYV team and are triaged with priority.

The Future Of Fitness Should Be Intelligent.
It Should Also Be Human.

STRYV is building AI-powered competition around a simple principle: technology should strengthen human effort, protect the player and earn trust through action.