AI Safety Roadmap: Urgent Action Needed Now

World leaders co-leading the UN's AI governance dialogue are urging immediate action on safety, following warnings from industry CEOs and financial regulators about risks including cyberattacks and bioterrorism. The call follows a July UN Global Dialogue in Geneva where 170 countries agreed on the need for common rules, including independent testing of new AI systems and mandatory incident reporting. The leaders propose three immediate steps: connecting scientific evidence to policy, establishing a safety baseline from existing frameworks, and funding inclusive participation. They recommend creating a trust fund for broader involvement and cross-border regulatory sandboxes, similar to financial sector testing. Specific goals are set for the next Dialogue in May, including an evidence assessment and a first pilot project.