Practical Resources for Intelligent Systems Management — a curated collection of frameworks, tools, regulations, papers, and open-source projects for responsible and trustworthy AI deployment in regulated industries.
Maintained by Sima Bagheri · LinkedIn · Medium
Focus areas: Enterprise AI governance · LLM deployment safety · Risk management · Regulatory compliance (NIST AI RMF, EU AI Act, ISO 42001) · Release readiness · Incident response
- Regulatory Frameworks
- Risk Management Frameworks
- Governance Tools & Platforms
- AI Testing & Evaluation
- Incident Management
- Model Cards & Documentation
- Academic Papers
- Datasets & Benchmarks
- Communities & Organizations
- Courses & Learning
- My Open-Source Frameworks
- NIST AI Risk Management Framework (AI RMF 1.0) — The U.S. government's voluntary framework for managing risks in the design, development, deployment, and use of AI systems. Organized around four functions: Govern, Map, Measure, Manage.
- NIST AI RMF Playbook — Practical guidance for implementing the AI RMF, with suggested actions for each subcategory.
- Executive Order on Safe, Secure, and Trustworthy AI — U.S. Executive Order (Oct 2023) establishing new standards for AI safety and security.
- NIST AI Safety Institute (AISI) — Federal body coordinating AI safety research and standards.
- OMB AI Governance Policy M-24-10 — Governance and risk management requirements for federal agency AI use.
- EU AI Act — The world's first comprehensive legal framework for AI, using a risk-based tiered approach (unacceptable, high, limited, minimal risk).
- EU AI Act Summary — Plain-language guide to the EU AI Act provisions and timelines.
- GDPR & AI — European Data Protection Board guidance on AI and GDPR intersection.
- ISO/IEC 42001:2023 — International standard for AI management systems. Provides requirements and guidance for establishing, implementing, maintaining, and improving an AI management system.
- ISO/IEC 23894:2023 — Guidance on risk management for AI systems.
- IEEE 7000 Series — IEEE standards for ethically aligned AI design.
- OECD AI Principles — International principles on trustworthy AI adopted by 46 countries.
- NIST AI RMF Core — Interactive version of the AI RMF with searchable categories and subcategories.
- Microsoft Responsible AI Standard — Microsoft's internal responsible AI framework, publicly shared.
- Google PAIR Guidebook — People + AI Research guidebook for designing human-centered AI.
- IBM AI Fairness 360 — Open-source toolkit for examining, reporting, and mitigating discrimination in ML models.
- MITRE ATLAS — Adversarial Threat Landscape for AI Systems — knowledge base of AI-specific adversarial tactics.
- OWASP Top 10 for LLMs — The 10 most critical security risks for LLM applications.
- Microsoft Responsible AI Toolbox
— Integrated suite for responsible AI assessment including error analysis, fairness, causal inference, and counterfactual analysis.
- Giskard
— Open-source AI quality testing platform for detecting biases, vulnerabilities, and performance issues.
- verifywise
— AI compliance platform with direct NIST AI RMF and EU AI Act mappings.
- Evidently AI
— Evaluate, test, and monitor ML and LLM models in production.
- WhyLabs — AI observability platform for model monitoring and drift detection.
- Fiddler AI — Explainable AI and model performance monitoring for enterprises.
- Microsoft PyRIT
— Python Risk Identification Toolkit for generative AI red teaming.
- LangFuse
— Open-source LLM observability and analytics.
- Holistic Evaluation of Language Models (HELM) — Stanford's comprehensive LLM evaluation framework across scenarios, metrics, and models.
- EleutherAI LM Evaluation Harness
— Unified framework for evaluating language models across 200+ tasks.
- DeepEval
— LLM evaluation framework with metrics for RAG, hallucination, and safety.
- TruLens
— Evaluation and tracking for LLM-based applications.
- RAGAS
— Evaluation framework for Retrieval Augmented Generation pipelines.
- MLflow Model Evaluation — Built-in model evaluation with support for LLMs and custom metrics.
- AI Incident Database — Crowdsourced database of AI incidents and failures across industries.
- AI Vulnerability Database (AVID) — Taxonomy of AI failure modes, biases, and vulnerabilities.
- NIST AI Incident Tracking — NIST guidance on AI incident classification and response.
- Weights & Biases Incident Retrospectives — Real-world ML incident retrospectives from practitioners.
- Model Cards for Model Reporting (Google) — Original paper introducing model cards as a transparency mechanism.
- Hugging Face Model Card Toolkit — Standardized model card format with template and auto-generation.
- Google Model Card Toolkit — Python toolkit for generating model cards programmatically.
- Datasheets for Datasets — Framework for documenting datasets with provenance, composition, and intended use.
- Concrete Problems in AI Safety (Amodei et al., 2016) — Foundational paper defining five practical AI safety problems.
- Stochastic Parrots (Bender et al., 2021) — Seminal paper on risks of large language models.
- Model Cards for Model Reporting (Mitchell et al., 2019) — Introduced model cards as a documentation standard.
- The Alignment Problem (Krakovna et al., 2020) — Survey of specification gaming in AI systems.
- Trustworthy AI (Varshney, 2022) — Practical guide to building trustworthy ML systems.
- Governing AI for Humanity (UN Advisory Body, 2024) — UN report on global AI governance frameworks.
- BigBench
— Collaborative benchmark for large language model evaluation beyond current capabilities.
- TruthfulQA — Benchmark measuring whether LLMs generate truthful answers.
- HarmBench — Standardized evaluation framework for automated red teaming.
- MMLU — Massive Multitask Language Understanding benchmark across 57 subjects.
- Partnership on AI — Multi-stakeholder organization advancing responsible AI practices.
- MLCommons — Open engineering consortium for ML benchmarks and safety evaluations.
- Montreal AI Ethics Institute (MAIEI) — Research institute for AI ethics with practitioner community.
- Center for AI Safety (CAIS) — Research organization focused on reducing societal risks from AI.
- FINOS (Fintech Open Source Foundation) — AI readiness resources for financial services industry.
- NIST National AI Initiative — U.S. government AI standards and research coordination.
- Future of Life Institute — Research on existential and catastrophic AI risks.
- Responsible AI practices (Google) — Google's practical guidance on responsible AI development.
- AI Ethics (fast.ai) — Free course on AI ethics and data ethics.
- Trustworthy AI (IBM) — IBM's trustworthy AI foundations certification.
- NIST AI RMF Workshop Videos — Free workshop recordings on implementing the AI RMF.
- Human-Centered AI (Stanford HAI) — Stanford's human-centered AI educational resources.
Frameworks I have built for AI governance and release readiness in regulated industries:
| Repository | Description | Stars |
|---|---|---|
| governance-playbook | End-to-end AI governance playbook aligned with NIST AI RMF |  |
| release-checklist | Risk-tiered release gate checklist for LLM/ML deployments with airc CLI |
 |
| nist-rmf-guide | Practitioner guide to implementing NIST AI RMF in regulated industries |  |
| release-governance | 5-stage release lifecycle framework with governance gates |  |
| accountability-patterns | Design patterns for human accountability in AI systems |  |
| regulated-ai | Starter kit for deploying AI in regulated industries (healthcare, finance, insurance) |  |
| multi-agent-governance | Governance framework for multi-agent AI systems |  |
| agent-eval | Evaluation framework for AI agents across correctness, safety, and reliability |  |
Contributions are welcome! Please read the Contributing Guidelines and open an issue before submitting a PR.
How to add a resource:
- Verify the resource is publicly accessible and actively maintained
- Add it to the appropriate section with a one-line description
- For GitHub repos: add a stars badge using
 - Open a PR with the title
Add: [Resource Name]
To the extent possible under law, Sima Bagheri has waived all copyright and related or neighboring rights to this work.