Google DeepMind describes world models as:
“AI systems that can use their understanding of the world to simulate aspects of it.”
— Google DeepMind, Genie 3: A new frontier for world models
| Paper | Year / Status | Security relevance | Link |
|---|---|---|---|
| World Models — Ha & Schmidhuber | 2018 | Foundational latent world-model architecture | Paper |
| Mastering Diverse Domains through World Models (DreamerV3) — Hafner et al. | 2023 | Planning and policy learning from imagined futures | Paper |
| Security of World-Model-Based Embodied AI: A Lifecycle of Threats, Defenses, and Evaluation — Liu et al. | 2026 · Preprint | Lifecycle threat model: poisoning, backdoors, spoofing, trajectory and supply-chain attacks | Paper |
| False Prophets: On the Security of World Models in Agentic Systems — Imgrund et al. | 2026 · ACM AISEC | Security impact of incorrect or manipulated world-model predictions on agents | Paper |
| TRAP: Tail-aware Ranking Attack for World-Model Planning — Duan et al. | 2026 · ACM Multimedia | Attacks trajectory ranking inside world-model-based planning | Paper |
| BadWorld: Adversarial Attacks on World Models — Shen et al. | 2026 · Preprint | Adversarial attacks against visual world-model rollouts | Paper · Code |
| When World Models Dream Wrong: Physical-Conditioned Adversarial Attacks against World Models — Guo et al. | 2026 · Preprint | Manipulates physical-condition inputs and degrades downstream perception/planning | Paper |
| When the World Lies: Backdoor Attacks on Latent World Models for Downstream Control — Riaño et al. | 2026 · Preprint | Backdoored world-model checkpoints and downstream control / supply-chain risk | Paper |