A recent assessment of global practices in frontier AI finds significant gaps in governance and outlines practical ways to close them. The authors review existing international frameworks, drawing on the United Nations recommendations, the European Union AI Act, NIST’s AI Risk Management Framework, and China’s AI Safety Governance Framework. By combining academic and industry safety studies, public incident data, and expert reviews, they map risks and weaknesses in current oversight. The report deliberately excludes military applications; it focuses on large‑scale, non‑military AI use. The aim is to improve preparedness and collective response to AI risks, advancing safe and reliable AI for humanity. These findings provide a reference for government bodies, international organizations, businesses, research institutions and other stakeholders seeking concrete governance mechanisms.