AI Safety Teams Struggle to Keep Up With Rapid Advances in Frontier Models

Teetering between breakthrough and breakdown, AI safety teams race to contain frontier models that evolve faster than their defenses—discover what fails next.