The big picture

The findings challenge industry claims that modern AI safety layers are robust. Guardrail removal appears accessible enough to lower the barrier for real-world misuse.