AI & Enterprise

How to optimise enterprise AI costs: Control compute structure first

Companies looking to reduce AI costs should do more than pick cheaper models, an argument says. They need to cut unnecessary tokens and computation themselves. Ben Canning (벤 캐닝), chief product officer at Alteryx, wrote that model costs are becoming a new burden as corporate AI use moves from experiments to large-scale operations. He said firms should first decide whether a large language model is needed and build business-logic workflows to avoid repeated processing.

AI & Enterprise

OpenAI blocks attempts to extract protected reasoning, cites links to Moonshot AI

OpenAI said it detected and blocked an organised effort to extract non-public reasoning information from its AI models at scale. The activity began on July 1 and surged on July 24-25, when 16,000 requests showing similar extraction patterns came from more than 4,000 users. OpenAI said related activity was confirmed across more than 15,000 users and was blocked by July 28. It described the method as adversarial distillation and said the core group was linked to people related to Chinese startup Moonshot AI.