Security & Privacy
SAFETY.md · 7 documents
Validation-Based Refusal: A Transparent Alternative to Pre-emptive Pattern Matching in AI Safety Systems
Proposes a validation-based refusal architecture that evaluates requests against explicit ethical axioms before deciding, reducing false positives by 60-80% while maintaining security.
Safety (Failsafe) Configuration
Documents PX4 failsafe actions, battery/RC/geofence loss triggers, and QGroundControl safety setup for drone recovery.
Safety (Failsafe) Configuration
Documents PX4 failsafe triggers and actions for battery, RC loss, data link loss, geofence, position loss, and offboard loss.
Safety
Explains why Rafx is an unsafe GPU API and how to manage resource lifetimes and state transitions manually or via the framework's render graph.
IntentShield
Pre-execution safety layer that audits AI agent actions against immutable rules before they run, blocking shell commands, file deletions, and jailbreaks.
Safety (Failsafe) Configuration
Documents PX4's failsafe system: battery, RC loss, data link loss, geofence, position loss, and offboard loss, with configurable actions and QGC setup.
Safety
Surveys LLM safety risks, government and private sector guidance, and safety benchmarks for developers and researchers.