Frontier AI Developers Urge Slowdown, but Enterprises Still Run Two-Year-Old Models

The gap between frontier labs calling for restraint and enterprises deploying well-understood, already-commercialized models is the real story here: safety debates at the research edge may have limited bearing on how AI is actually used in business today. If enterprises remain content with older, proven models, pressure for slower development could prove less costly commercially than it sounds. Whether that lag persists as commercialized capabilities advance is the open question.