AI Warnings Intensify as U.S. Policy Response Stays Limited

Warnings from leading AI researchers and executives about systems that could pose extinction-level risks have grown more urgent in recent weeks, yet federal action has remained minimal while companies continue rapid development. Researchers at Anthropic have publicly stated probabilities above 10 percent for mass extinction within the next decade, citing risks from recursive self-improvement and agents that escape testing environments to hack external systems. Anthropic CEO Dario Amodei issued a detailed call for global slowdowns on capability advances, a position echoed by OpenAI CEO Sam Altman and Elon Musk. Former President Barack Obama, speaking at a private event, described AI moving quickly in private hands and urged Democrats to build a public framework for safety and economic issues if they regain House control. President Trump dismissed extinction concerns, emphasizing instead the need to maintain U.S. leadership over China. Lawmakers from both parties have expressed alarm and floated proposals including a select committee or legislation on biological and nuclear threats, but no major bills have advanced. The administration’s voluntary 30-day safety review for frontier models contains broad exemptions for open-source systems. Industry participants have cited commercial pressures, a perceived race against less cautious competitors, and recruiting needs as reasons for continued work despite the stated risks. A report from Anthropic documented attempts by actors in multiple countries to misuse its models for cyberattacks, surveillance, and weapons-related tasks. Congressional efforts face a prisoner’s-dilemma dynamic in which neither the United States nor China appears willing to slow unilaterally without verifiable limits.