Tag: agent safety
Browse our exclusive articles!
Human-Guided Harm Recovery for Safer AI Agents
Explore human-guided harm recovery methods that help AI agents safely recover from harmful actions, ensuring responsible computer system use.
Agent Safety Risks: How Benign Instructions Hide Vulnerabilities
Discover how harmless user instructions expose critical vulnerabilities in computer-use agents and why new safety benchmarks like OS-BLIND matter.
HINTBench: Benchmark for Intrinsic Agent Risk Trajectories
Explore HINTBench, a benchmark for detecting intrinsic risks in agent trajectories, improving safety in autonomous systems with advanced risk auditing.