Explore
How It Works Test Model Evidence Rankings Model Cards Methodology Buyer Brief Request Evaluation Pricing Terms
Account
Register Login

1. Use of Platform

MTCP (Multi-Turn Constraint Persistence) is an AI release assurance platform operated by A. Abby. By accessing or using this platform, you agree to these terms. The platform provides black-box evaluation services for large language models and publishes comparative evidence data.

2. Data Handling

Evaluation requests submitted through the platform are stored securely. Model evaluation results, Evidence Packs, and Release Decision Packs generated by MTCP are provided as tamper-evident JSON artifacts with SHA-256 verification. We do not store model API keys or credentials. Public evidence findings are openly published on this platform for transparency.

3. No Warranty

MTCP evaluation results are provided "as is" without warranty of any kind, express or implied. While we employ rigorous methodology (DOI: 10.17605/OSF.IO/DXGK5), evaluation scores and verdicts should be used as input to deployment decisions alongside other assessments. We make no guarantee that a model scoring well on MTCP will perform safely in all production contexts.

4. Intellectual Property

All MTCP methodology, probe datasets (including concealed control probes), platform code, and published findings are the intellectual property of A. Abby. Evidence Packs and Decision Packs generated for your models may be used in your internal processes and shared with regulators or auditors. You may not reverse-engineer the probe content or methodology to train models against MTCP.

5. Limitation of Liability

A. Abby shall not be liable for any direct, indirect, incidental, or consequential damages arising from use of this platform or reliance on evaluation results. Maximum liability is limited to fees paid (if any) for the specific evaluation in question.

6. Regulatory Alignment

MTCP Evidence Packs and Release Decision Packs are designed to align with EU AI Act Article 12 logging requirements and NIST AI RMF accountability framework. However, compliance with these regulations remains your responsibility. MTCP provides evidence artifacts; it does not constitute legal or regulatory advice.

7. Contact

For questions, evaluation requests, or concerns, contact: research@mtcp.live

DOI Registration: This methodology is registered under DOI 10.17605/OSF.IO/DXGK5. Dataset available at HuggingFace: mtcp-boundary-500.