Vercel’s analysis underscores a critical vulnerability in AI infrastructure: the high cost of inference makes endpoints lucrative targets for theft. While their proposed solution, BotID, appears promising, it may not fully address the underlying economic incentives driving these attacks. As AI models grow more expensive to run, the industry could face increasing pressure to adopt stricter access controls or shift toward more closed ecosystems. This tension between openness and security may become a defining challenge for AI providers.