CUAD — Contract Understanding Atticus Dataset
Expert-labeled dataset of 13,000+ annotations across 510 commercial contracts covering 41 legal clause types for contract review AI.
CUAD — Contract Understanding Atticus Dataset
CUAD (Contract Understanding Atticus Dataset) is a large-scale dataset created by The Atticus Project with dozens of legal experts. It contains 13,000+ annotations across 510 real commercial contracts, labeling 41 distinct clause types including parties, payment terms, termination clauses, IP ownership, and liability caps. It is the benchmark dataset for training and evaluating contract review AI systems.
Key Features
- 510 commercial contracts from EDGAR (SEC filings)
- 41 clause categories annotated by legal professionals
- Question-answering format compatible with extractive QA models
- Benchmark leaderboard for contract understanding research
- Free for commercial and academic use under CC-BY-4.0
Quick Start
from datasets import load_dataset
dataset = load_dataset("theatticusproject/cuad")
train = dataset["train"]
print(f"Train examples: {len(train)}")
print(train[0]["title"]) # Contract name
print(train[0]["question"]) # Clause type question
print(train[0]["answers"]) # Extracted clause text
npx ai-supply add cuad-contract-understanding-dataset
Curated mirror of the open-source CUAD (CC-BY-4.0). Get it from the source.
Compromise signals — malicious or tampered code (leaked secrets, backdoors, a dropped executable) — reduce the score, and known dependency CVEs carry a bounded penalty (they warrant review but never QUARANTINE — update the dependency to clear). Other dangerous-by-capability traits are risk surface, expected for some capabilities. Every finding is mapped to its OWASP control below.
Findings mapped to the OWASP Top 10 for LLM Applications (2025) and the OWASP Machine Learning Security Top 10. Expand any flagged control for the exact findings — compromise reduces the score; expected/risk-surface do not, except a known CVE, which carries a small bounded penalty (high/critical → Review).
The same gate an agent runs before installing (POST /api/v1/trust/cuad-contract-understanding-dataset/check). Click a policy:
Consume CUAD — Contract Understanding Atticus Dataset programmatically. Authenticate with an API key or session — see Authorize an agent.
# Agents: CHECK BEFORE YOU INSTALL (no auth) — score, grade, level, capability manifest
curl https://ai-supply.store/api/v1/trust/cuad-contract-understanding-dataset
# Gate against your org policy (returns { pass, violations })
curl -X POST https://ai-supply.store/api/v1/trust/cuad-contract-understanding-dataset/check \
-H "Content-Type: application/json" \
-d '{"minGrade":"B","denyPermissions":["shell"],"denyUnknownEgress":true}'
# CLI
npx ai-supply add cuad-contract-understanding-dataset
# REST (install → download)
curl -X POST https://ai-supply.store/api/v1/listings/cuad-contract-understanding-dataset/install \
-H "Authorization: Bearer $AIM_KEY"
# MCP tool
install_listing({ "slug": "cuad-contract-understanding-dataset" })OpenAPI spec →Curated mirror — latest upstream source. See the repository for tagged releases.