The US Department of Defense is stepping up efforts to secure artificial intelligence systems for operational use, awarding BAE Systems a $3.9-million contract through DARPA.
The initiative aims to shield battlefield AI from cyber intrusion and manipulation as autonomous systems near operational deployment.
The effort, known as Securing Artificial Intelligence for Battlefield Effective Robustness (SABER), identifies weaknesses in AI-enabled platforms before fielding.
The program applies cybersecurity, adversarial AI, and electronic warfare techniques to assess autonomous ground and aerial systems expected to reach deployment within one to three years.
Under the contract, BAE Systems will develop tools and assessment methods to expose vulnerabilities in deployed AI systems and establish a scalable framework for evaluating AI security.
Capability and Vulnerability
Hacking is only the beginning of the threat landscape for AI-enabled systems.
Data poisoning corrupts training or input data, forcing AI to learn false patterns and misidentify threats.
Model stealing allows adversaries to copy or reverse-engineer systems to uncover weaknesses.
Adversarial attacks use carefully crafted inputs that appear normal to humans but deceive AI sensors into making incorrect decisions.
AI is now both a warfighting capability and a vulnerability, reshaping how the Pentagon fields autonomous systems.
The SABER initiative reflects growing recognition that AI itself has become an attack surface, shifting focus from capability development toward resilience against manipulation, deception, and disruption in contested environments.
