RiskLine

Problem

Engineering teams inventorying AI systems for the EU AI Act need classification that runs in CI, stays offline, and shows which article fired — not a portal that guesses with an LLM.

Who it's for

Platform engineers, compliance-aware developers, and design partners building AI governance into pipelines — before buying a full GRC suite.

How it works

1. Describe the AI system (purpose, data, context, autonomy, affected population).
2. Engine matches against a versioned EU AI Act ruleset.
3. Get tier + clause refs + rationale + recommended controls + disclaimer.

Features

• Deterministic classification — versioned, auditable JSON ruleset
• Offline-first CLI — no data leaves your machine
• HTTP API with OpenAPI contract
• Single Go binary — CI-friendly (OPA/Trivy distribution model)
• Apache 2.0 open source

Quick start

go install github.com/new-world-coder/riskline/cmd/riskline-cli@v0.1.0-alpha

riskline-cli examples/curl/hiring-assist.yaml
riskline-cli --json examples/curl/hiring-assist.yaml

Sample output

Hiring Assist → high_risk · Annex III (4)(a) recruitment

System:     Hiring Assist
Risk tier:  high_risk
Matched:    high-risk-recruitment → Annex III (4)(a)

Controls:
  - Risk management system and data governance
  - Human oversight of ranking/filtering
  - Logs and technical documentation
  - Bias and discrimination assessment

Status & disclaimer

v0.1.0-alpha — pre-1.0. EU AI Act only. Partial Annex III coverage.

This classification is an advisory tool based on a versioned ruleset. It is not legal advice and is not a substitute for qualified counsel or a formal conformity assessment.

github.com/new-world-coder/riskline