deepseek

China's banks are accelerating efforts to deploy DeepSeek artificial intelligence (AI) models in multiple business scenarios,MKsport with major state-owned and joint-stock banks integrating them into intelligent customer service, operational efficiency and credit approval.

The country's biggest commercial lender, Industrial and Commercial Bank of China (ICBC), recently completed the localized deployment of the latest open-source DeepSeek AI model, the bank announced on Saturday. The model has been integrated into ICBC's large model matrix framework to enhance intelligence across financial business scenarios. 

According to the bank's statement, ICBC has built a multi-layered framework with multi-modal perception and multi-task collaboration, forming a collaborative application matrix of over 10 large models and more than 2,000 traditional models, enabling optimized coordination among different models.

The bank has already implemented large language models to empower over 20 major business areas, with more than 200 real-world applications, showcasing the strong driving force of AI technology in promoting high-quality development of the financial industry.

In terms of product innovation, ICBC has developed the ChatDealing intelligent dialogue system, reshaping the trading process and significantly increasing transaction volumes.

Centered on the whole credit process, ICBC has developed a dedicated credit approval risk control assistant, which integrates credit policy inquiries, report writing, risk assessment, data analyses and loan approval recommendations. 

Also, in February, the Postal Savings Bank of China announced that, leveraging its own large model, the bank has deployed and integrated the DeepSeek-V3 model and the lightweight DeepSeek-R1 inference model, as reported by the Xinhua Net.

The bank has applied the DeepSeek into its AI Assistant, adding logical reasoning capabilities to enhance service precision. Through functions such as deep analyses, it accurately identifies user's needs and provides personalized and scenario-based service solutions. Leveraging efficient reasoning performance, the bank has accelerated response time and task processing efficiency.

Among joint-stock banks, the Shanghai Pudong Development Bank has also deployed the DeepSeek-R1 671B large model, which is embedded in digital assistant applications, enhancing scenarios like intelligent Q&A, financial analyses and report writing, as reported by CNR.

The move by multiple banks to deploy DeepSeek AI models has positive effects on banking operations and digital transformation, such as improving business efficiency and service quality, Wang Peng, an associate research fellow at the Beijing Academy of Social Sciences, told the Global Times on Thursday.

Wang added that the introduction of DeepSeek models can enhance risk management capabilities and drive business innovation by improving the efficiency of decision-making.

However, given the high sensitivity of bank customers' information and DeepSeek's reliance on large volumes of data for training, the risk of data leakage increases, Wang said. As AI technologies are widely adopted, banks should ensure that the development, training and application of DeepSeek models comply with relevant laws, regulations and supervisory requirements.

Additionally, in order to enhance the model's interpretability, banks should develop visualization tools to present the decision-making process and rationale of the model in an intuitive way, Wang said. 

He also noted that banks need to strengthen internal talent training by developing systematic programs to improve staff's technical capabilities to keep pace with the speed and efficiency of model iteration.