AI accuracy notice

Version 2026-06-12

ARK's outputs are AI-generated predictions, expressed as confidence ratios — not statements of fact.

• Every annotation, threat level, role, and event is a probabilistic estimate. ARK can and does produce false positives and false negatives.
• Outputs must not be the sole basis for any safety, security, legal, employment, insurance, or law-enforcement decision. A qualified human must review before action.
• ARK does not identify individuals by name and is not a biometric identity system; person labels (e.g. "person 5, striped shirt") are scene-local descriptors, not identities.
• Accuracy varies with video quality, lighting, occlusion, viewing angle, and scene complexity. Confidence scores are model estimates, not guarantees.
• ARK is decision support. You remain the data controller and the decision-maker, responsible for lawful, fair, and human-reviewed use.

This is the same class of notice used by major AI and analytics platforms: the system is a powerful assistant that can be wrong, and its output is a starting point for human judgment, not a verdict. The short form of this notice ships in the disclaimer field of every API response.

Honour Systems Inc. · See also the Acceptable Use Policy.