Tester Agreement
Governing your participation as a benchmark tester
Last Updated: December 18, 2025
Why This Matters: The integrity of the Great Commission Benchmark depends on maintaining the confidentiality of test questions. If questions become publicly available or are shared with AI model providers, they may be incorporated into training data, rendering our benchmark ineffective at providing accurate, unbiased evaluations.
Introduction and Purpose
Why this agreement exists
This Tester Agreement governs your participation as a tester in the Great Commission Benchmark. By running benchmark tests through the Service, you agree to be bound by this Agreement in addition to our Terms of Service and Privacy Policy.
This Agreement establishes your obligations to protect the confidentiality of benchmark materials and outlines the consequences of violating these terms.
Key Definitions
Important terms used throughout this agreement
- "Confidential Information"
- All benchmark test questions, prompts, scenarios, evaluation criteria, expected responses, scoring rubrics, and any other materials used in the testing process that are not publicly available on our leaderboards or website.
- "Test Questions"
- The specific questions, prompts, and scenarios presented to AI models during benchmark testing.
- "Model Provider"
- Any company, organization, or individual that creates, trains, fine-tunes, or distributes AI models, including but not limited to OpenAI, Anthropic, Google, Meta, Mistral, and their employees, contractors, or affiliates.
- "Public Disclosure"
- Sharing, publishing, posting, or otherwise making information available to any third party, whether through social media, websites, forums, academic papers, presentations, or any other medium.
- "Training Use"
- Using information to train, fine-tune, improve, or otherwise enhance any AI model or system.