This comprehensive survey examines the critical factors undermining trustworthiness in AI systems, focusing on failure modes, vulnerabilities, and biases.

• Analyzes three key dimensions: safety alignment, privacy protection, and bias mitigation
• Addresses specific concerns in large language models, including harmful content generation
• Explores advanced techniques for identifying and preventing privacy attacks
• Provides a structured framework for evaluating AI trustworthiness

For security professionals, this research offers valuable insights into identifying vulnerabilities and implementing protective measures against emerging AI threats, establishing standards for responsible AI development.

Trustworthy AI on Safety, Bias, and Privacy: A Survey