AI-generated abuse content surged in 2025, with watchdogs warning that the technology is making harmful material easier to create and spread.