White House Imposes Pre-Release Vetting on OpenAI’s GPT-5.6 Launch

Washington’s gate-kept debut of GPT-5.6 echoes the 1990s U.S. export controls on high-grade encryption—when PGP was classed as ‘munition’ until controls relaxed in 2000—illustrating how states react when civilian code gains strategic potency. The move fits a century-long pattern of governments asserting jurisdiction over dual-use technologies (radio in 1912, nuclear reactors after 1945, gene-editing tools post-2012). By inserting itself between Silicon Valley and end-users, the U.S. signals that cutting-edge general-purpose AI is migrating from consumer software to a regulated strategic asset, with national-security logic overriding laissez-faire norms championed just a decade ago. Whether this moment becomes a lasting regime—embedding AI into export-control architectures like ITAR—or a brief precaution hinges on if near-term reviews actually mitigate cyber risk without throttling innovation. On a 100-year horizon, the precedent may mark the start of sovereign licensing of cognition-scale computation, much as states today license radio spectrum or nuclear fissile material; if so, June 2026 could be remembered as the point when the free-flow software era definitively ended for frontier AI.