| Mobile Web

Kakao unveils AI personalisation and safety design, memory limited to consented services

Kakao unveiled a personalisation and safety design that separates how user memory is used from the authority given to AI agents to act. It said memory will be provided only within the scope of user consent and only to the requesting service or agent. Kakao also outlined safety checks that verify not just final answers but execution steps, including action plans and tool calls. It introduced an operating strategy that splits roles between on-device and server models.