We have some massive structural problems with AI that nobody seems willing to address. 1. AI companies are positioned to replace their own customers. Businesses increasingly need AI to compete, but every prompt potentially exposes their workflows, expertise, and intellectual property. AI companies claim policies protect this data, but what stops them from repurposing it for training under another name? You're effectively helping build your replacement. Worse, AI providers control the models, have access to capabilities you don't, and can spend $20M subsidizing a competing product until you're bankrupt. 2. China threatens the entire American AI business model. Anthropic accused three Chinese labs of using 16 million Claude exchanges to distill its models.