Hong Kong to Launch HKGAI V1: Local AI Chatbot Powered by DeepSeek Set for 2025 Release
Image Credit: Simon Zhu | Splash
Hong Kong is set to introduce HKGAI V1, a locally developed artificial intelligence chatbot, in the second half of 2025. The chatbot leverages technology from Chinese AI startup DeepSeek and is designed to deliver instant responses tailored to the city’s linguistic and cultural context. The initiative, announced by Secretary for Innovation, Technology and Industry Sun Dong during a government media tour in April 2025, marks a significant milestone in Hong Kong’s efforts to establish itself as a regional leader in AI.
Background: Advancing Hong Kong’s AI Capabilities
HKGAI V1 is being developed by the Hong Kong Generative AI Research and Development Centre (HKGAI), part of the government’s InnoHK research program. The chatbot’s purpose is to support a range of everyday tasks for users in Hong Kong, such as generating travel recommendations, meeting summaries, and multimedia content—functions comparable to those of leading AI chatbots like OpenAI’s ChatGPT. A key feature is its ability to operate fluently in Cantonese, Putonghua, and English, increasing its accessibility for Hong Kong’s diverse population.
DeepSeek, based in Hangzhou and founded in 2023 by Liang Wenfeng, has quickly gained recognition for its cost-efficient large language models (LLMs). Its recent models, such as DeepSeek-V3, have drawn global interest for delivering performance rivalling leading Western systems, reportedly achieving similar results at a fraction of the training cost. While industry estimates suggest DeepSeek’s latest model was trained for about US$5.6 million, compared to much higher figures for Western models, official costs for GPT-4 remain undisclosed.
Development and Features
HKGAI V1 builds on DeepSeek’s advanced large language models and utilizes a full-parameter fine-tuning approach, aiming for fast and contextually relevant responses. The user interface is designed to be intuitive, with special attention to Hong Kong-specific knowledge and multi-language support. More than 70 government departments have already begun trialing the chatbot, indicating strong institutional backing for the project.
The underlying technology uses Nvidia H800 GPUs—currently among the most advanced chips available to Chinese developers due to export controls. DeepSeek’s models employ technical strategies such as mixed-precision arithmetic and optimized load-balancing to maximize efficiency and reduce computational costs while maintaining strong performance. The Hong Kong University of Science and Technology (HKUST), a key partner, hosted a global launch event for the project in February 2025, underscoring the collaboration between academia, government, and industry.
Once the government trials are complete, HKGAI V1 is expected to become available to the private sector and the general public, with the potential for adoption beyond Hong Kong.
Opportunities and Advantages
HKGAI V1 offers several benefits for Hong Kong’s public sector, businesses, and residents. The use of DeepSeek’s models makes the technology more cost-effective, facilitating wider deployment. Support for Cantonese and Hong Kong-specific knowledge allows the chatbot to provide more relevant and localized responses, increasing its practical value for users. DeepSeek’s open-source licensing (MIT License) also gives developers in Hong Kong the flexibility to further customize and innovate on top of the platform.
Strategically, the project supports Hong Kong’s goal of becoming a technology hub within the Greater Bay Area, in alignment with national initiatives to promote technological self-reliance. By showcasing competitive AI capabilities, Hong Kong aims to enhance its profile as a leader in regional innovation.
Challenges and Concerns
Despite these advantages, HKGAI V1 faces scrutiny over security and privacy. While Hong Kong officials have emphasized the implementation of strict data protection measures, concerns have been raised—particularly in U.S. congressional reports—about the potential for data access by mainland Chinese entities, given DeepSeek’s origins. In April 2025, a U.S. House committee identified DeepSeek as a possible national security risk, although no formal restrictions have yet been imposed by Hong Kong authorities.
Additionally, relying on a single external provider for core AI infrastructure could create long-term dependency risks. As AI adoption accelerates, there is also uncertainty about the potential impact on local employment, especially in industries susceptible to automation, though specific effects in Hong Kong remain to be studied.
Global Restrictions and Reliability Concerns
DeepSeek's AI chatbot technology has faced scrutiny and restrictions in multiple countries due to privacy and security concerns. Italy was the first country to ban DeepSeek, following an investigation by the Italian Data Protection Authority. Australia, South Korea, Taiwan, and Canada have also implemented bans or restrictions on the use of DeepSeek on government devices. In the United States, several federal agencies, including the Navy and NASA, have prohibited the use of DeepSeek, and legislation has been proposed to ban its use on all federal devices.
In addition to these restrictions, DeepSeek's chatbot has been criticized for its reliability. A study conducted by NewsGuard revealed that DeepSeek's chatbot failed to provide accurate responses to news-related prompts 83% of the time, ranking it among the least reliable AI models tested. The chatbot was found to repeat false claims 30% of the time and provide vague or unhelpful responses in 53% of cases.
TheDayAfterAI News
We are a leading AI-focused digital news platform, combining AI-generated reporting with human editorial oversight. By aggregating and synthesizing the latest developments in AI — spanning innovation, technology, ethics, policy and business — we deliver timely, accurate and thought-provoking content.