Tag: agentic AI safety
Browse our exclusive articles!
Green Shielding: Enhancing Trustworthy AI with User Focus
Discover Green Shielding, a user-centric method improving AI reliability and safety in healthcare by addressing input variations in large language models.
Layerwise Convergence Fingerprints for LLM Misbehavior Detection
Discover Layerwise Convergence Fingerprinting, a tuning-free method to detect runtime misbehavior in large language models with high accuracy and security.
Human Feedback for Semantic Skill Discovery in AI
Discover how human feedback enhances semantic skill discovery in AI, ensuring diverse, safe, and value-aligned behaviors through SRSD.
Jailbreaking Frontier AI Models via Intention Deception
Explore how multi-turn intention deception exploits vulnerabilities in frontier AI models like GPT-5, revealing critical safety risks and para-jailbreaking...
Effective Prompt Injection Defenses for Large Language Models
Discover how output filtering secures large language models from prompt injection attacks and protects sensitive data effectively.