Cody Wild

About

Cody Wild is a researcher whose work sits at the intersection of artificial intelligence safety and reinforcement learning. Her most influential contribution is the landmark paper "Adversarial Policies: Attacking Deep Reinforcement Learning" (2019), which has garnered 92 citations and fundamentally reshaped how the field thinks about security in multi-agent systems. In this work, Wild demonstrated that an adversarial agent could learn a policy that, while appearing benign, subtly manipulates the environment to cause a victim agent to fail—without ever directly tampering with the victim's observations. This insight revealed a critical blind spot in deep RL: even when an attacker cannot directly modify another agent's inputs, they can still exploit learned vulnerabilities through indirect means. The paper has become essential reading for researchers working on robust and safe AI, and it has spurred a new line of inquiry into adversarial dynamics in multi-agent settings. Wild's work stands as a clear warning that as we deploy RL agents in the real world, we must account for adversaries who can learn to exploit the very structure of interaction itself.

Research Focus

Key Achievements

Top Papers

1. 1
   Adversarial Policies: Attacking Deep Reinforcement Learning

   92 citations · 2019

Key Collaborators

Contact & Links