US Drops AI Safety Reviews, Prioritizing Speed Over Caution

The Trump administration’s abrupt reversal on a planned executive order requiring pre-release government safety reviews for frontier AI models marks a pivotal policy shift. This move, framed as necessary for maintaining competitive dominance over China, effectively abandons a cautionary, risk-mitigation framework in favor of prioritizing developmental speed. This decision lands squarely in the ongoing conflict between AI accelerationists and safety proponents, signaling that the US government now views geopolitical competition, much like the space race, as the primary lens for AI strategy, sidelining existential risk concerns for now. The administration’s action fundamentally alters the risk-reward calculation for leading AI labs by eliminating a significant regulatory hurdle. The winners are clearly defined: Google, Microsoft-backed OpenAI, and Anthropic are now unshackled from a process that would have slowed model deployment and increased R&D costs, creating an asymmetric advantage against international competitors. The losers are AI safety advocates and the public, as the guardrails against the release of dangerously capable or biased systems are dismantled, placing the onus of responsibility almost entirely on corporate self-governance. The long-term trajectory this sets is one of escalating capabilities and systemic risk. Within months, expect a flurry of more powerful model releases and intensified lobbying to prevent future regulatory attempts. Over the next 1-2 years, this will likely create a schism between the US 'move fast' approach and the EU's precautionary AI Act, forcing a global bifurcation in AI governance. The critical variable is whether corporate self-policing can prevent a major AI-driven incident before a public or state-level backlash forces a much harsher, reactive regulatory crackdown.