Responsible AI

Why this matters here

VogicAI builds systems that identify people, monitor public space and inform operational decisions. That capability carries obligations. This page sets out the principles we apply to how our products are designed and deployed.

Human oversight by default

VogicAI systems are decision-support systems. A match, alert or classification is an input to a trained operator's judgement, not a substitute for it. We design for a human in the loop on consequential actions, and we do not market our systems as fully autonomous.

Lawful and authorised deployment

Facial recognition, crowd monitoring and aerial surveillance are subject to legal and procedural controls that differ by jurisdiction and by sector. We expect every deployment to operate under the customer's lawful authority, within its own governance framework, and with the approvals that framework requires.

Honest capability claims

We describe what our systems are designed to do under stated conditions, and we state the factors that affect real-world performance. We do not claim guaranteed accuracy, zero false positives, or performance that has not been measured.

Where a customer needs assurance for a specific environment, we prefer evaluation on their own footage and conditions over a generic benchmark.

Accuracy, bias and evaluation

Recognition performance varies with image quality, illumination, pose and the composition of reference data. We test across the conditions our customers actually operate in, and we support customer-run acceptance testing before an operational go-live.

Data minimisation and control

Our deployment models exist so that sensitive operational data does not have to leave customer infrastructure. Retention windows, access scopes and watchlist governance are configured by the customer.

• Role-based access control
• Audit logs for identity searches and watchlist changes
• Configurable retention and deletion
• Edge, on-premise and air-gapped operation

Raising a concern

If you believe a VogicAI system is being used in a way inconsistent with these principles, write to [email protected]. Concerns are reviewed by the engineering and leadership teams.