RiskForge

EU AI Act Article 9 risk management, as a developer workflow.

RiskForge is an open-source CLI that turns EU AI Act Article 9 compliance from a consultant invoice into a 30-minute developer workflow.

Answer 37 guided questions across 8 EU AI Act risk dimensions. RiskForge produces a tamper-evident Risk Management File (JSON, PDF, or Markdown) carrying a SHA-256 self-verifying digest, suitable for inclusion in your Annex IV technical documentation pack and ready for your legal team and downstream compliance toolchain. (Not a substitute for notified-body conformity assessment.)

Built by AI Exponent LLC. Apache 2.0. Runs entirely offline after pip install.

Documentation: the full documentation covers installation, the command reference with exit codes, the EU AI Act Article 9 mapping, CI integration, plugin authoring, the audit-chain design, and an FAQ.

Quick Start

pip install riskforge

# 1. Register your AI system
riskforge init \
  --name "Loan Scoring Model" \
  --sys-version "2.1" \
  --purpose "Automated credit scoring for retail loan applications." \
  --provider "Acme Financial Services" \
  --category essential_services

# 2. Record your Article 6(2) Annex III self-classification (required before export)
riskforge system classify <system-id> --confirm

# 3. Run the guided 8-dimension risk assessment (~30 minutes of thinking)
riskforge assess <system-id> \
  --assessor-name "Alice Chen" \
  --assessor-role "AI Governance Lead"

# For CI or reproducible fixtures, run it non-interactively from a YAML answers file:
#   riskforge assess <system-id> -a "Alice Chen" -r "AI Governance Lead" --answers answers.yaml

# 4. (Optional) Record mitigations and accept residual risk
riskforge risk mitigate <system-id> <risk-id> \
  -m "Remove postcode feature; add demographic parity monitoring" \
  -c preventive --owner "ML Platform" --residual-likelihood 2 --residual-severity 3
riskforge risk accept <system-id> <risk-id> --rationale "Residual within appetite after controls."

# 5. (Optional) Derive Article 9(6)-(8) test requirements per open or knowledge-gap risk
riskforge tests generate <system-id>

# 6. Check completeness before export (exit 1 if a FAIL gate is unmet)
riskforge validate <system-id>

# 7. Export your Article 9 Risk Management File
riskforge export <system-id> --format pdf --output rmf.pdf
riskforge export <system-id> --format json --output rmf.json

# 8. Verify integrity anytime (exit 2 if the file was tampered)
riskforge verify --file rmf.json

Why RiskForge

EU AI Act Article 9 requires providers of high-risk AI systems to maintain a documented risk management system throughout the system's lifecycle.

The current alternatives:

^{¹ Indicative market figures gathered from public Big-4 governance-engagement quotes and 2024–2026 enterprise GRC pricing pages. Not a benchmark study; your mileage will vary by scope, jurisdiction, and incumbent advisor.}

Architecture

RiskForge has four strictly-decoupled layers with CI-enforced import boundaries:

graph TD
    CLI["CLI (Typer)<br/>riskforge init / assess / validate / export / verify"]
    Engine["Engine Layer<br/>AuditEngine · RiskEngine · ValidateEngine<br/>AssessEngine · ExportEngine · TestDerivationEngine"]
    Storage["Storage (FileStore)<br/>YAML + JSONL · chmod 600/700 · async · pluggable ABC"]
    Adapters["Integration Adapters<br/>RAGBenchmarkingAdapter · TraceForgeAdapter<br/>Discovered via Python entry_points"]

    CLI -->|"calls engine functions"| Engine
    Engine -->|"reads/writes via StorageBackend ABC"| Storage
    Engine -->|"adapter pattern, no hard imports"| Adapters

    style CLI fill:#1e3a5f,color:#fff
    style Engine fill:#1e3a5f,color:#fff
    style Storage fill:#1e3a5f,color:#fff
    style Adapters fill:#1e3a5f,color:#fff

State on disk:

your-project/
├── riskforge.yaml                     project manifest (chmod 600)
├── .riskforge/                        (chmod 700)
│   ├── audit.jsonl                    append-only hash-chained audit log
│   ├── audit.lock                     serialises audit appends
│   ├── .nodelete                      deletion sentinel
│   └── systems/<system-id>/
│       ├── system.yaml
│       ├── register.yaml
│       └── mitigations.yaml
└── rmf-<id>-<export>.json             exports land where --output says (default: here)

Plain YAML plus JSONL: readable by regulators without RiskForge installed, and diff-able in GitHub PRs.

AI Exponent compliance toolchain, planned integration

RiskForge is designed to integrate with the broader AI Exponent toolchain. Today, only RiskForge and rag-benchmarking are available on PyPI. The other nodes below are on the public roadmap and will integrate via plain JSON files when they ship.

graph LR
    RAG["rag-benchmarking<br/>(accuracy evidence)<br/><i>shipped</i>"]
    TF["TraceForge<br/>(data governance)<br/><i>roadmap</i>"]
    RF["RiskForge<br/>(Art. 9 RMS)<br/><i>shipped</i>"]
    TD["TransparencyDeck<br/>(Art. 13 docs)<br/><i>roadmap</i>"]
    CB["ConformityBot<br/>(Art. 43 cert)<br/><i>roadmap</i>"]
    CCO["Compliance Officer<br/>(PDF)"]

    RAG -->|"benchmark_report.json"| RF
    TF  -.->|"trace_report.json (planned)"| RF
    RF  -->|"rmf.pdf"| CCO
    RF  -.->|"rmf.json (planned)"| TD
    RF  -.->|"rmf.json (planned)"| CB

    style RF fill:#c9a84c,color:#000,stroke:#c9a84c
    style RAG fill:#1e3a5f,color:#fff
    style TF fill:#6b7685,color:#fff,stroke-dasharray:5
    style TD fill:#6b7685,color:#fff,stroke-dasharray:5
    style CB fill:#6b7685,color:#fff,stroke-dasharray:5
    style CCO fill:#2d5a2d,color:#fff

All current connections are plain JSON files on disk. RiskForge never calls external APIs.

EU AI Act Article 9 Coverage

graph LR
    A9_1["Art. 9(1)<br/>Establish RMS"] --> REG["Register lifecycle<br/>Version history<br/>Audit log"]
    A9_2a["Art. 9(2)(a)<br/>Identify risks"] --> QB["Guided question bank<br/>8 dimensions · 37 questions"]
    A9_2b["Art. 9(2)(b)<br/>Estimate misuse risks"] --> PAT["Risk patterns<br/>6 Annex III scenarios"]
    A9_4["Art. 9(4)<br/>Risk measures"] --> MIT["Mitigation docs<br/>Vague-detection"]
    A9_6_8["Art. 9(6)–(8)<br/>Testing requirements"] --> TEST["riskforge tests generate<br/>Per-risk metric hints"]
    A9_9["Art. 9(9)<br/>Vulnerable groups"] --> VG["Dedicated questions<br/>Mandatory flag"]
    A9_10["Art. 9(10)<br/>Documentation"] --> AUD["Append-only JSONL<br/>SHA-256 hash chain"]

    style A9_1 fill:#1e3a5f,color:#fff
    style A9_2a fill:#1e3a5f,color:#fff
    style A9_2b fill:#1e3a5f,color:#fff
    style A9_4 fill:#1e3a5f,color:#fff
    style A9_6_8 fill:#1e3a5f,color:#fff
    style A9_9 fill:#1e3a5f,color:#fff
    style A9_10 fill:#1e3a5f,color:#fff

Cross-maps to: NIST AI RMF (GOVERN/MAP/MEASURE/MANAGE) · ISO/IEC 42001 (Clauses 6.1, 8.4, A.6–A.9) · Colorado AI Act (SB 24-205, reset by SB 26-189) · Texas HB 149 (TRAIGA)

Disclaimer: RiskForge produces documented evidence for Article 9 compliance. It does not substitute for qualified legal counsel or notified body conformity assessment.

Validation Gates

Before every export, riskforge validate runs 8 gates:

Features

Contributing

The easiest contribution requires zero Python, edit a YAML file and open a PR.

Add a question to an existing dimension:

# src/riskforge/_data/question_bank/privacy.yaml
- id: PRIV-007
  text: "Does the system process special category data under GDPR Article 9?"
  guidance: "Special category data includes health, biometric, racial, or political data."
  annex_iii_categories: [essential_services, employment]
  default_likelihood_hint: 3
  default_severity_hint: 5
  article_refs: ["Art.9(2)(a)", "Art.10(3)"]
  nist_rmf_ref: "MAP 1.5"
  iso42001_ref: "Clause A.7"
  regulatory_status: settled

Add a risk pattern, edit src/riskforge/_data/patterns/patterns.yaml.

Fix a bug or add a feature, see CONTRIBUTING.md.

git clone https://github.com/aiexponent/riskforge
cd riskforge
make dev-setup   # pip install -e ".[dev]" + pre-commit install
make test        # 57 tests, all must pass
make lint        # ruff check + format

Privacy

RiskForge makes zero outbound network connections in CLI mode, enforced in CI with pytest-socket --disable-socket.

RiskForge v1.1.3 | Apache 2.0 | Zero telemetry | aiexponent.com

Your AI system's risk data never leaves your machine unless you explicitly deploy the optional API server (pip install riskforge[server]). That server is experimental: it is not security-hardened and is not part of the flagship test suite, so run it only on a trusted, local network.

Releases

License

Apache 2.0, free to use, modify, and distribute.

Built by AI Exponent LLC, hello@aiexponent.com

Part of the AiExponent open-source AI governance toolchain:
litmusai (Art. 5) · license-compliance-checker (Art. 53) · rag-benchmarking (Art. 15) · riskforge (Art. 9) · agentic-document-analyser (Art. 9 / Annex IV)