A third element is needed in the U.S. AI strategy to enhance safety: developing systems that actively counter malicious applications, particularly AI capable of self-defense.