Global AI regulation shifts from voluntary principles to mandatory government-led safety testing. The US and UK are instituting pre-release national security assessments for major AI models from firms like OpenAI and Google. International cooperation, such as the UK-Australia pact, aims to align safety standards. Compliance and proven safety are becoming key competitive advantages, moving AI governance into an evidence-based implementation phase.