AI policy is increasingly driven by national security risks, leading to reactive and unpredictable government measures. The development and release of advanced AI models like Anthropic’s Mythos and Fable 5 have triggered U.S. government interventions, including export controls and pressure on companies like Meta and OpenAI to delay releases or submit models for review. This shift signifies a new era where AI is treated as a potential national security threat, with the U.S. government acting as a de facto gatekeeper for public access to AI models.

• National security risks are now the primary factor shaping AI policy, overshadowing other concerns.
• Recent events involving advanced AI models have led to unexpected and aggressive policy responses.
• The U.S. government has implemented measures such as export controls and voluntary review frameworks for AI developers.
• The release of Anthropic’s Mythos and Fable 5 models prompted swift government action due to identified vulnerabilities.
• President Trump signed an Executive Order promoting a voluntary framework for AI developers to engage with the government on security.
• A memorandum was released to accelerate AI adoption within the U.S. military.
• The U.S. government has intervened in the release of AI models, citing national security concerns.
• There is a lawsuit challenging the government’s authority to force AI companies to shut down model access.
• The U.S. government is reportedly pressuring major tech companies to submit AI models for review.
• A new normal for U.S. AI policy has emerged, characterized by a ‘state of exception’ and swift, decisive action.
• The U.S. government is becoming the de facto entity deciding public access to AI models.
• The U.S. announced “Pax Silica,” a U.S.-led alliance aimed at countering China’s dominance in AI.
• China’s Premier expressed concerns about the speed of technological progress and the risk of losing control.
• Future AI policy is expected to be unpredictable, with new AI releases treated as potential national security threats.