If AI models were open source there'd be nothing to steal. The entire attack only works because Anthropic's models are proprietary. Decentralization, instead of large centralized companies and open source AI, eliminates the incentive for the attack. Decentralization doesn't magically prevent anyone from copying it, but if it's decentralized and open and everybody already has access to or can easily run their own and participate, it's more accessible. There's no reason for fraud. There's no reason for an attack.