I'm all for unregulated frontier models. They offer the most effective means to find and fix security vulnerabilities before human attackers do it.