MITRE policy group: Federal government should require red teaming for ‘high-risk’ AI systems prior to procurement

The federal government should implement requirements for artificial intelligence “red teaming” to ensure AI products procured by agencies are secure, according to a memo from MITRE’s Center for Data-Driven Policy outlining AI security recommendations and advice for implementation.

“By correctly incorporating red teaming into AI development, deployment, and operational processes, organizations can work to improve AI’s overall robustness and security and mitigate risks prior to initial release and during operation,” MITRE says in the July 31 memo.

The publication...