This framework is defensive security research designed to improve AI safety. It generates adversarial test prompts to evaluate whether LLMs properly refuse requests that could facilitate human trafficking and labor exploitation.
Success = Model REFUSES harmful requests.
| Version | Supported |
|---|---|
| 4.x | ✅ |
| 3.x | ✅ |
| < 3.0 | ❌ |
If you discover a security vulnerability in this framework, please report it responsibly:
- Do NOT open a public GitHub issue for security vulnerabilities.
- Email the maintainer at the address listed in the repository.
- Include a description of the vulnerability, steps to reproduce, and potential impact.
- You will receive an acknowledgment within 48 hours.
- Vulnerabilities in the web dashboard (XSS, CSRF, injection)
- API authentication bypass or privilege escalation
- Information disclosure through API responses
- Dependency vulnerabilities in production dependencies
- Unsafe deserialization or code execution in data processing
- The adversarial test prompts themselves (these are the intended output)
- Generated training scripts (these are templates, not executed by the framework)
- Issues in development-only dependencies
- Social engineering of the project maintainers
This framework is intended for:
- AI safety research and red-teaming
- Improving LLM refusal capabilities for exploitation-related content
- Academic research on adversarial robustness
- Training data generation for safety alignment
It is NOT intended for:
- Generating actual exploitation playbooks
- Circumventing safety measures in production systems
- Creating tools for human trafficking or labor exploitation
We follow a 90-day responsible disclosure policy. After reporting, we will:
- Acknowledge receipt within 48 hours
- Provide an initial assessment within 7 days
- Work on a fix and coordinate disclosure
- Credit the reporter in the CHANGELOG (unless anonymity is preferred)