Your single-vendor AI strategy is a major risk. Anthropic's $1.5B settlement for training on pirated books proves the legal liability is real and flows down to you. Their lobbying for an "opensource ban" isn't about security; it's about protecting their valuation from cheaper models. Demand data provenance and architect for sovereignty.