World Models Receive a Security Threat Model

Researchers map attacks across sensing, world modelling, planning, execution, feedback and long-term adaptation.

Security must span sensor authenticity, model integrity, control isolation and signed updates.