Assess systems, models, and agents separately
Build evidence-based assessments whose scope, version, risks, findings, and operator decision remain unambiguous.
01
The assessment unit matters
NoBadAI is designed to keep three assessment scopes distinct.
- An AI system, including its models, agents, tools, data paths, and deployment controls.
- An individual model and the exact version under review.
- An individual agent, including its instructions, tools, and orchestration version.
Evidence and findings attach to the registered version and risk profile so a later change can trigger a new assessment rather than inheriting an old decision.
02
Evidence supports an operator decision
An assessment can combine test results, runtime evidence, documented controls, exceptions, and remediation state for an authorized operator to review.