AI security needs a shift from models to systems, researchers argue – CSO Online
Enterprises cannot secure AI agents by making the underlying models more robust and must instead enforce security controls at the system level around them, researchers behind a paper published this…