Evaluations
How Cyberdroid AI is measured on an organization's own telemetry.
How an evaluation works
Agree on representative cases and success measures before testing. The evaluation should show whether the system helps the team complete the work it will actually use.
Measures
Detection is compared with the current process on replay and representative telemetry. AI Investigation includes difficult cases, conflicting evidence and missing data, with success and stop criteria agreed in advance.
Detection
The scorecard tracks accuracy, false alarms, time to a finding and the hardware used.
Investigation
The scorecard counts right and wrong conclusions and incomplete cases, plus minutes per case and cost per accepted investigation.
Limits
Scores
Interpret a score alongside the test data, labels and operating conditions. Those details explain what the result supports and where further evaluation is needed.
Outcomes
Measure useful findings, review effort and operating cost in the proposed environment. Include the work needed to correct or complete a case.
Status
Cyberdroid AI Detection is available now to select organizations. Cyberdroid AI Investigation is in early access, so its evaluation runs as a design partnership.
Standards
Naming a clause is not a claim of certification.
NIST
AI RMF MEASURE 2.1 asks for documented test sets and metrics, and MEASURE 2.5 for limits on generalizing.
EU
Articles 13 and 15 of the EU AI Act ask for declared accuracy metrics. The scorecard declares them first.
