Anthropic Project Glasswing Targets Safer AI Agents Before They Go Rogue - HotHardware