The Trump administration’s tightening grip on frontier AI model releases is creating a crisis that threatens the entire AI industry, not just individual companies. Following the withdrawal of Anthropic’s Mythos and Fable models from public availability, OpenAI’s GPT-5.6 has now been placed in a government-gated limited preview, with access approved on a customer-by-customer basis before any general release can proceed.
CEO Sam Altman reportedly suggested the preview period could last only a couple of weeks, but Anthropic’s Mythos has already been in restricted preview for months with no general release in sight. Even brief delays carry serious economic consequences for AI labs burning significant capital on model development and data centre infrastructure, at a time when the industry is under intense pressure to demonstrate financial returns.
Former White House AI adviser Dean Ball, soon to join OpenAI, has argued that the current process amounts to an involuntary licensing regime for frontier AI, made worse by the absence of clearly defined safety standards or articulated risks that regulators are actually trying to mitigate. Without that framework, the government lacks both the expertise and capacity to conduct meaningful model evaluations, creating the conditions for indefinite delays.
Crucially, OpenAI and Anthropic now face identical constraints and identical risks. Industry observers have framed this as a competitive battle, with accusations of regulatory capture on one side and political cozying on the other. But the consequences of a dysfunctional approval process fall on the entire sector equally, from model developers to data centre investors to enterprise customers.
Addressing the underlying concerns around cybersecurity, biorisk, and AI alignment will require the industry to pursue collective solutions rather than use safety and regulation as competitive leverage — a significant cultural shift for a sector accustomed to zero-sum rivalry.
