Tag: agentic AI safety
Browse our exclusive articles!
Enhancing Multilingual AI Safety with Self-Distillation
Discover how self-distillation improves safety alignment in multilingual AI, boosting protection in low-resource languages without extensive data.
RouteHijack: Exploiting Routing Vulnerabilities in MoE LLMs
Discover how RouteHijack targets routing in Mixture-of-Experts LLMs, revealing critical vulnerabilities and advancing AI safety research.
EvoJail: Adaptive Diverse Jailbreak Prompts for LLMs
Discover EvoJail, an evolutionary framework generating diverse, adaptive jailbreak prompts to enhance safety and robustness in large language models.
Fixing Safety Failures in Agentic AI Guard Models
Explore vulnerabilities in agentic AI guard models and how Fisher-Weighted Safety Subspace Regularization restores safety and reliability.
Safety in Embodied AI: Risks, Attacks & Defenses Survey
Explore key risks, attacks, and defense strategies in embodied AI to ensure safe, reliable human-agent interactions and robust real-world deployment.