Classify risk, surface gaps, export in one pass

Upload model cards, architecture notes, or policy excerpts — Agent Mai maps your narrative to EU AI Act expectations, estimates risk tier, and returns a remediation-ready report your legal and engineering leads can share.

Ingest

PDF, JSON, paste

Classify

Annex III signals

Review

Gap + priority

Export

JSON for GRC

What happens in the audit run

Designed for speed without hiding assumptions — every output cites the evidence slice it relied on.

Rapid ingestion

Drag files or paste raw text. We normalize heterogeneous docs so reviewers are not blocked by formatting.

Risk & Annex signals

Heuristic alignment to high-risk categories and documentation depth checks against Annex IV-style completeness.

Dual runtime

Switch between Mai Cloud for velocity or Vault mode when your policy forbids external inference.

Actionable gaps

Each finding includes severity, suggested owner discipline, and draft language you can edit before submission.

Engineering-friendly

Outputs reference technical artifacts — data lineage hints, evaluation caveats, and monitoring hooks.

Next steps

Jump to the integrity report, Annex IV composer, or prohibited-practice scan from the same navigation shell.

Audit sequence

IngestClassifyReviewExport

Inference mode

Upload or paste technical documents

PDF, JSON model cards, dataset manifests — combine with pasted text below.

Select file

Annex IV composer

Synthesize draft technical documentation aligned with Annex IV sections before you file with authorities.

Open workflow

API key (preview)

Wire CI/CD for automated monitoring — coming soon.

am_live_****************827a

Regulatory context

Interpretation targets EU AI Act Art. 11 and Annex IV documentation completeness — always confirm with counsel.

Continue to your integrity report

After the run, share the JSON or PDF with security and legal, then track remediation in your existing PM tools.

View report workflow Run safety scan