A universal "protocol" could be settled to address this issue. Official institutional AI agents (e.g., Gov bots, banks, stores...) could use this protocol to guarantee a good AI output without compromising user's privacy, ensuring global-level accountability over the system. Sharing datasets and training knowledge between independent AI agents across multiple contexts could be done using blockchain protocols anonymizing users ID. These AI systems should submit themselves to communicate with other agents using this protocol. An international committee would be responsible for managing and ensuring rules to operate, evolve, and promote the treaty as a primary industry standard.