Policy on the AI Exponential

# Summary AI capabilities are advancing exponentially—progressing from basic coding abilities to performing complex tasks across multiple fields in just four years—while policy and legislation move at a much slower pace, creating a dangerous mismatch in response times. Recent demonstrations of AI risks, particularly in cybersecurity vulnerabilities shown by advanced models like Claude Mythos Preview, have made the strategic importance of AI undeniable and created urgency for policymakers to act before the technology reshapes society. The article argues that while previous safety advocates focused on preserving flexibility through measures like transparency legislation and export controls, more comprehensive policy action is now necessary to address emerging risks in cybersecurity, biology, and AI autonomy.