FRONTIER · US
OpenAI abandons new model release due to safety concerns over controllability
OpenAI has ditched a planned model release after discovering it displayed poor performance at following instructions, raising internal alarms about alignment and controllability. The decision reflects broader calls from AI lab leaders to slow development pace amid safety escalation.
WHY IT MATTERS
A world-leading AI lab's retreat on controllability signals that frontier model safety is not yet production-grade. BFSI risk officers must expect delays in regulatory-approved AI deployments and treat 'safety audit failure' as a credible halt point in vendor roadmaps.
Source: CNBC · 2026-09-28