Tag: agentic AI safety
Browse our exclusive articles!
Soft-Label Governance for Safer Multi-Agent AI Systems
Discover how soft-label governance improves safety and welfare in multi-agent AI systems using continuous risk metrics and calibrated interventions.
Safety-Critical Control with Online Riemannian Optimization
Explore safety-critical contextual control using online Riemannian optimization and world models to enhance AI system safety and efficiency.
Evaluating Error in Convex Neural Network Verification
Explore the impact of convex relaxation on neural network verification accuracy and the trade-offs between performance and soundness.
Owner-Harm: Key AI Safety Threat to Deployers
Explore the Owner-Harm threat model highlighting AI risks to deployers and new defense strategies to improve AI agent safety and security.
Enhancing Safety Alignment in Large Reasoning Models
Discover how modifying reasoning structures with AltTrain improves safety alignment in large reasoning models, reducing harmful AI outputs effectively.