AI's New Gatekeepers: How Security and Economics Are Reshaping Frontier Model Access

Major AI developers like Anthropic and OpenAI are restricting access to their most powerful frontier models, such as Mythos and gpt-5.5-cyber, to a select group of U.S.-based companies. This marks a critical shift away from the prevailing assumption that cutting-edge AI capabilities would remain widely available, potentially widening the gap between leading U.S. tech firms and the rest of the world. As compute resources tighten, security concerns mount, and government oversight expands, the future of AI innovation appears to be moving toward selective distribution rather than broad accessibility.