The U.S. Defense Advanced Research Projects Agency (DARPA) is seeking input from industry developers and academic institutions on current and emerging technologies for assessing vulnerabilities in AI-enabled systems.

DARPA aims to identify techniques and tools that: 

• Address diverse adversarial access threat models (white, grey, black, hidden box)

• Evaluate vulnerabilities across the entire AI system pipeline; and

• Account for platform-specific factors like environmental conditions, multi-modal sensors, and system purpose.

“This Request for Information (RFI) seeks information on the current state of art assessment techniques and tools for testing (i.e., red teaming)

AI-enabled battlefield (cyber-physical) systems deployed in tactical, operational, and strategic operating environments,” DARPA said in a document attached to a notice posted on SAM.gov. 

Responses are invited from all qualified sources, including private and public companies, individuals, universities, university-affiliated research centers, nonprofit research institutions, and U.S. government-sponsored laboratories.