New mandatory guardrails will apply to AI models in high-risk settings, with businesses encouraged to adopt new safety standards starting now.