Safety & Ethics — Free AI Tools

Bias detection, guardrails, explainability, evaluation.

• IBM AI Fairness 360 — 70+ fairness metrics + 10 mitigation algorithms
• NeMo Guardrails — Programmable LLM guardrails — PII, jailbreak, fact-check
• DeepEval — 30+ metrics + red teaming 40+ vulnerabilities
• SHAP — Gold standard feature attribution — 20K+ stars, industry standard
• lm-eval-harness — 60+ benchmarks — powers the Open LLM Leaderboard
• Ragas — RAG evaluation — faithfulness, precision, recall metrics
• Llama Guard — Meta's open-source content safety classifier — 8B, runs locally
• Promptfoo — Test and evaluate LLM prompts — catch regressions before production