Invent ways to evade AI safety rules

Back to top button