OpenAI says the changes to its model training will increase compute overhead by 20% of observed inference workload; the increase will not be handed to customers (Thomas Claburn/The Register)

OpenAI announced that recent adjustments to its model-training pipeline will add compute overhead equal to roughly 20% of the current inference workload, and the additional cost will be absorbed internally rather than billed to customers. The company also said it has tightened safety protocols and paused reinforcement-learning training for two weeks following the Hugging Face breach and related security concerns.