What happened
Maeil Business Newspaper reports that Kakao unveiled Kanana Space, a voice-oriented AI assistant designed to operate inside KakaoTalk. The assistant is intended to understand conversational context, retrieve relevant past activity and coordinate Kakao services, while leaving final approval to the user. Kakao plans to recruit closed-beta testers in December and launch the service in the first quarter of next year, according to the report.
Maeil Business Newspaper reports that Kakao presented Kanana Space at an event identified in the article as “If Kakao 26.” The assistant is described as appearing within KakaoTalk, including as a floating interface over an active chat, and as being summonable by swiping the screen with two fingers. The report says users will be able to speak with the assistant and convert the conversation to text when needed.
The reported use cases include checking a schedule while chatting, finding a product from a previous order and applying a coupon before payment. In the example given by the outlet, the user still makes the final approval and decision. Kakao also described a four-level model system spanning , small language models, flagship models and frontier models. An AI “conductor” is reported to select models and tools and determine the order of work based on each request.
According to Maeil Business Newspaper, Kakao will accept pre-applications through the Kanana Space website and select some applicants for a closed beta in December. The report says Kakao plans a launch in the first quarter of next year. It does not provide a specific launch date, pricing, final list or confirmation that the service will be generally available at launch.
Why it matters
Kakao is positioning AI as an integrated layer across a widely used communications platform rather than as a separate chatbot. If the reported design works as intended, users could ask for actions in ordinary conversation and receive personalized results based on prior interactions or purchases. That could make AI more practically useful, but it also raises important questions about consent, privacy, accuracy and how much personal context the system can access. The report does not independently confirm performance, availability or safeguards.
The product’s significance lies in the proposed integration of AI with KakaoTalk’s existing communication and service ecosystem. Rather than requiring users to open a separate assistant, Kanana Space would interpret requests in the context of ongoing conversations and past activity. That model could reduce the effort needed to find information or initiate routine tasks, especially when the user’s request is incomplete or expressed conversationally.
Personalization also creates risks that are not resolved by the announcement. Access to past conversations, purchases and schedules could expose sensitive information or produce results that users do not expect. The reported requirement for final user approval is relevant for purchases, but the source does not explain how consent, data access, retention, correction of mistakes or escalation for ambiguous requests will work. No independent testing or user results are provided.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
A route planner searches possible journeys using explicit rules. What does this illustrate about AI?
What to watch next
The immediate milestones are Kakao’s December closed beta and the planned first-quarter launch. Key unknowns include the exact launch date, supported services, eligible users, pricing, geographic availability, device requirements, data-retention practices and the controls users will have over conversation history and personalization. It will also be important to see whether Kanana Space can reliably complete multi-step tasks without making unauthorized purchases or other consequential changes.
The closed beta should clarify who can use Kanana Space, what KakaoTalk data it can access and which services it can operate. It should also reveal whether the assistant is limited to recommendations and or can actually execute reservations, orders, remittances and other actions.
Practical evaluation should focus on reliability and user control: whether the system accurately recalls prior activity, asks for clarification when context is uncertain, applies the correct discounts, prevents unintended transactions and clearly shows what information or tools it used. Pricing, supported devices, regional availability and privacy settings remain unknown from the report.