HUMA-TECH8 min read
The HUMA Software Audit Playbook
Seven questions every responsible AI product should answer before applying for certification.
HUMA Certificate Team
TechAuditAI Ops
Is the core workflow deterministic?
If the main promise of your product collapses when an LLM is offline, you are not Level 1 or 2 material. Lock your primary flows to deterministic logic and relocate AI helpers to additive sidecar experiences.
Can users opt-out without losing value?
Optional Utility certification demands frictionless opt-outs. Show us the exact toggle, API flag, or environment variable that disables AI-generated results without degrading the baseline experience.
Where are the guardrails documented?
For Responsible Engine certification tier, we look for written evidence: abuse playbooks, rate-limit policies, model monitoring dashboards. Screenshots count. Vaporware policies do not.
