OpenAI pauses frontier RL training citing safety and alignment gaps

OpenAI CEO Sam Altman said the company is pausing some frontier reinforcement learning training because rapid model development is outstripping safety, alignment, security and monitoring standards. Builders watching: capability races are now explicitly bounded by risk infrastructure maturity—invest in safety testing as a feature gate, not an afterthought.

Source: Mint ↗