Back to evaluations
Public evaluation

RiskIdentification

The evaluation will assess the system's ability to accurately identify compliance risks in AI projects, provide justified recommendations, and generate comprehensive, well-structured compliance reports based on the given regulations and project details.

Evaluation type
task based
Challenge
Generative AI Regulatory Compliance with LangGraph and Claude Sonnet 4.5
Difficulty
Advanced
Rigor
Unspecified

Evaluation overview

How the linked challenge is judged: tasks, benchmarks, and criteria count.

Tasks
2
Benchmarks
0
Criteria
0

Task templates

Inputs and expected outputs.

Task 1

RiskIdentification

Evaluate the system's ability to identify relevant compliance risks based on a project description and specific regulatory text.

Input format

JSON object: {'project_description': 'string', 'regulation_text': 'string'}

Output format

JSON object: {'identified_risks': [{'risk_type': 'string', 'severity': 'low|medium|high', 'justification': 'string', 'relevant_clauses': ['string']}]}

Task 2

ComplianceReportGeneration

Assess the system's ability to generate a structured compliance report for a given AI project, including risks, recommendations, and evidence.

Input format

JSON object: {'project_name': 'string', 'project_description': 'string', 'identified_risks': [...]}

Output format

Markdown string of a compliance report, including summary, detailed risks, and mitigation strategies.