This research introduces TrustGen, a systematic framework and benchmarking platform to assess and improve the trustworthiness of Generative Foundation Models.

• Provides a comprehensive analysis of global AI governance laws, policies, and industry standards
• Develops a multi-dimensional trustworthiness assessment framework covering safety, fairness, explainability, and more
• Implements an open-source benchmarking platform to evaluate GenFM trustworthiness
• Offers practical perspectives for future research and development

For security professionals, this research delivers essential tools to identify vulnerabilities, assess risks, and enhance the safety and reliability of foundation models in production environments.

On the Trustworthiness of Generative Foundation Models: Guideline, Assessment, and Perspective