Melvine’s AI Analysis #72 - The Role of Artificial Intelligence and Generative AI at ICBC

Oct 11

Melvine Manchau

September 10, 2025

The Industrial and Commercial Bank of China (ICBC), one of the world’s largest banks by assets, has been at the forefront of integrating artificial intelligence (AI) and generative artificial intelligence (Gen AI) into its operations. As the financial services sector undergoes a seismic transformation driven by emerging technologies, ICBC’s adoption of AI reflects a strategic commitment to enhancing efficiency, improving customer experiences, and staying competitive in a rapidly evolving market. This note explores ICBC’s use cases and initiatives for AI and Gen AI, industry trends, competitor strategies, expected impacts, associated risks and challenges, and the regulatory environment shaping AI adoption in China’s banking sector. One of ICBC’s notable strides is in embracing open-source large language models and customizing them for banking.

The bank has localized deployment of the DeepSeek open-source LLM (a prominent Chinese generative AI model) and integrated it into a multi-layered “large model matrix” framework. This framework combines over 10 large models and 2,000+ traditional models in a collaborative system, enabling different AI models to work together across tasks. By 2024, ICBC had applied large language models to 20 major business areas with more than 200 real-world applications, underscoring the breadth of its AI rollout. For instance, ICBC developed “ChatDealing”, an intelligent dialogue system for trading desks that reshapes trade execution and has significantly boosted transaction volumes by automating deal-making conversations.

Similarly, ICBC built a credit approval risk-control assistant that streamlines the loan process – handling credit policy Q&A, writing credit reports, conducting risk assessments, and even generating loan approval recommendations to assist human officers. The bank has also introduced AI tools for internal efficiency, such as AI coding assistants and an investment research assistant for financial markets, as well as an internal audit assistant, to accelerate analysis and reporting tasks. In customer-facing services, ICBC has experimented with 3D digital humans (virtual customer service avatars named “Gong Xiao Zheng” and “Gong Xiao Cheng”) to enhance interactive experiences in digital banking. These initiatives demonstrate ICBC’s all-round approach – deploying generative AI and machine learning in everything from front-office customer engagement to middle-office decision support and back-office automation

Early AI Developments at ICBC

ICBC was an early mover in exploring AI technologies to improve banking services and internal processes. In the late 2010s, ICBC (Asia) experimented with biometric recognition and AI for personalized services and risk control. For example, it piloted VIP facial recognition to identify high-value clients in branches, and used big data and AI for tailored product recommendations and fraud detection.. By around 2018, ICBC (Asia) was planning a 24/7 intelligent chatbot capable of voice interactions in Mandarin, Cantonese, and English to handle customer inquiries, aiming to boost service efficiency and compliance in cross-border banking. These early initiatives laid the groundwork for ICBC’s AI capabilities, focusing on enhancing customer experience and strengthening risk management.

In the following years, ICBC expanded its AI efforts internally. By 2020, the bank had introduced an intelligent digital assistant named “Gong Xiao Zhi” (alias Gino) as a customer service robot, and implemented AI in compliance and anti-fraud systems. Notably, ICBC developed a voiceprint-recognition security measure for fraud prevention and launched the “ICBC Brains” intelligent anti-money laundering system covering end-to-end KYC, risk classification, and suspicious transaction monitoring. These early AI applications – from chatbots to anti-fraud “brain” systems – illustrate ICBC’s proactive approach to leveraging AI for both front-end services and back-end risk controls well before the generative AI boom.

ICBC’s AI and Gen AI Use Cases

ICBC’s commitment to AI has accelerated in recent years, especially with the advent of large language models (LLMs) and generative AI. The bank prides itself on leading the domestic banking industry in developing proprietary AI capabilities. In 2023, ICBC became the first Chinese bank to build its own large-model AI technology system, staying at the forefront of AI innovation in finance. Internally dubbed “Industrial Smart Ocean”, ICBC’s flagship AI model ecosystem has scaled up dramatically.

The bank reports deploying over 200 AI-powered use cases and logging more than 1 billion AI calls per year. These range from algorithmic credit advisors that help evaluate loans to forex trading assistants and intelligent risk detection systems in operations. The Industrial Smart Ocean model itself handles workloads that were once performed by 45,000 human employees, highlighting how extensively AI is being integrated to automate and augment ICBC’s workforce. ICBC has leveraged AI and Gen AI across various functions to optimize operations, enhance customer engagement, and strengthen risk management. Based on available insights, the following are key use cases:

Risk Management and Fraud Detection: ICBC employs AI algorithms to enhance its risk control processes, particularly in credit risk assessment and fraud detection. Machine learning models analyze vast datasets from customer transactions, credit histories, and external sources to identify patterns indicative of potential risks or fraudulent activities. For instance, AI algorithms can detect anomalies in real-time transaction data, enabling proactive measures to mitigate fraud. Gen AI further supports this by generating synthetic data for stress testing and scenario analysis, allowing ICBC to simulate potential risk scenarios without compromising sensitive customer information, as noted in recent industry reports. AI enhances ICBC’s risk monitoring through pattern recognition and predictive analytics. The bank’s “ICBC Brains” AML system, upgraded with AI, monitors transactions across the entire bank for signs of money laundering or fraud, automatically flagging anomalies in real time AI-driven fraud detection and cyber-risk models also guard digital channels. Notably, ICBC deployed a voiceprint recognition system to verify identities and detect fraudsters via voice analysis, boosting both security and customer convenience In credit risk, ICBC uses AI to detect early warning signs in its portfolio – a smart risk control platform provides forward-looking risk identification and alerts to manage problem loans proactively.
Credit and Lending: AI models are used as algorithmic loan officers, evaluating credit applications with greater speed and data-driven consistency. ICBC’s AI-based credit advisor tools analyze borrower data and give preliminary risk assessments, while the credit approval assistant automates large parts of the underwriting process (from pulling policy rules to drafting loan memos). These tools help reduce manual workload and error rates in lending, enabling faster loan decisions especially for small business and consumer loans.
Trading and Investment: In ICBC’s financial markets division, AI is transforming trading and research. The ChatDealing system allows traders to interact via natural language with a trading platform that can execute orders or retrieve market data on command, making trading more efficient. Meanwhile, investment research assistants powered by AI help analysts sift through market information, news, and financial reports to generate insights. ICBC’s AI can draft research summaries and even code quantitative models – Bank of China’s similar model generated over 13 million lines of code in one month for internal use – illustrating how AI accelerates development tasks.

Customer Service and Personalization: ICBC has integrated AI-powered chatbots and virtual assistants to streamline customer interactions. These systems, often powered by natural language processing (NLP) and Gen AI, handle routine inquiries, provide personalized financial advice, and assist with account management. Gen AI enables these chatbots to generate human-like responses, improving customer satisfaction and reducing the workload on human agents. For example, ICBC’s mobile banking app likely uses AI to offer tailored product recommendations based on customer behavior and preferences, aligning with its digital banking strategy.
Data Analytics and Predictive Modeling: ICBC utilizes big data and AI analytics to develop predictive models that optimize resource allocation and anticipate market trends. These models help the bank make data-driven decisions, such as identifying opportunities for cross-selling or adjusting investment strategies based on market conditions. Gen AI enhances this capability by generating insights from unstructured data, such as customer feedback or social media trends, as part of its focus on big data infrastructure. AI chatbots and digital assistants handle routine customer inquiries, provide 24/7 support, and personalize banking advice. ICBC’s “Gong Xiao Zhi” intelligent robot and 3D digital avatars serve customers on mobile apps and online channels, improving response times and consistency. Voice-command capabilities (in multiple languages) and facial recognition for VIP clients are used to make services more accessible and tailored.
Personalized Financial Products: AI and Gen AI are used to create customized financial products tailored to individual customer profiles. For instance, ICBC’s mobile banking app incorporates AI features for personalized financial advice, helping customers with investment strategies and financial planning. This personalization enhances customer loyalty and drives revenue growth, with reports indicating heavy investment in digital banking technologies for tailored financial products.
Internal Operations and Process Automation: A significant share of ICBC’s AI use cases are inward-facing, aiming at efficiency gains. The bank has implemented Robotic Process Automation (RPA) bots in combination with AI, creating an internal marketplace of automated routines for repetitive tasks (e.g. data entry for loan processing or regulatory report preparation). Moreover, ICBC’s generative AI models produce internal documents and code at scale – the bank notes that its AI systems can generate millions of lines of code and extensive internal reports, saving engineers and auditors enormous time. An example is the internal audit assistant which quickly reviews transactions and drafts audit findings using AI, allowing human auditors to focus on high-risk issues.

Collectively, these use cases show that ICBC is embedding AI deeply into its core processes, far beyond just chatbots. By leveraging AI for everything from customer chats to risk controls and coding, ICBC is transforming into what it calls a “Digital ICBC” (D-ICBC) or a “technology bank”, aligning with its strategic vision of technology-driven banking.

ICBC’s AI Initiatives

ICBC has pursued several strategic initiatives to integrate AI into its operations, reflecting its ambition to remain a leader in digital banking:

Investment in Digital Banking Technologies: ICBC has heavily invested in digital banking platforms, upgrading its mobile banking app with AI features for personalized financial advice. The bank plans to enhance its online service capabilities to provide seamless banking experiences, as part of its strategic objectives outlined in recent reports.
Collaboration with Technology Firms and Academic Institutions: ICBC actively collaborates with technology firms like Baidu, Tencent, and Huawei, and academic institutions to develop cutting-edge AI solutions. These partnerships facilitate the integration of advanced AI technologies into banking services and support talent development through training programs and research collaborations, emphasizing the importance of skilled professionals.
Development of AI-Powered Financial Products: ICBC is developing AI-driven financial products, such as personalized investment strategies and risk management tools. These products leverage AI to analyze customer data and market trends, offering tailored solutions to meet individual needs, as part of its focus on enhancing customer experience.
Focus on Sustainable and Responsible AI: ICBC emphasizes sustainable AI practices, minimizing environmental impact and ensuring ethical AI use. The bank invests in robust governance frameworks to balance innovation with societal values, recognizing the need for regulatory compliance and ethical considerations.
Exploration of Quantum Computing and AI Synergy: ICBC is exploring the potential of quantum computing to enhance AI capabilities, particularly for complex data processing and advanced analytics, as part of its long-term innovation strategy.

Industry Trends: AI in Chinese Banking

ICBC’s AI push is part of a broader wave of AI adoption across China’s banking industry. Following the breakthrough of models like OpenAI’s ChatGPT, Chinese banks – especially the large state-owned banks – have raced to develop or deploy their own AI models.

By late 2024, ten of China’s top banks (including all the “Big Six” state banks and major joint-stock banks such as Ping An Bank and China Merchants Bank) had rolled out generative AI in hundreds of operational scenarios. In fact, at least 11 large banks launched AI model projects after ChatGPT’s debut, integrating AI across functions like office automation, customer service, marketing, investment research, and risk control. This industry-wide trend marks a shift from pilots and chatbots to full integration of AI into critical workflows.

A key enabler of this trend has been the emergence of domestic large language models such as DeepSeek. DeepSeek is an open-source Chinese LLM that many banks have adopted and fine-tuned for internal use.

For example, China Construction Bank (CCB) privately deployed a large financial model based on DeepSeek R1 in 2023 to power its internal services.

Bank of China (BoC) not only employed DeepSeek R1 for tasks like automated coding and report generation (with one report noting it produced 13.37 million lines of code in a month), but BoC also announced a sweeping AI Development Action Plan and invested heavily in AI research and deployment.

China Construction Bank built a comprehensive AI-as-a-Service platform anchoring hundreds of use cases – as of 2024, CCB had 168 AI use cases in production, with 7,000 model deployments covering 193 scenarios, and it reported that 16 core banking processes now rely on AI tools.

Postal Savings Bank of China, another state-owned lender, developed its own in-house large model suite called “Youzhi”; by 2024, it had generated over 1.1 million lines of code with AI assistance and cut document generation time by 90% through these tools. Even some joint-stock banks have joined in: Shanghai Pudong Development Bank (SPD), for instance, deployed the DeepSeek-R1 67-billion parameter model within its digital assistant, enhancing capabilities in intelligent Q&A, financial analysis, and report writing.

This widespread adoption is redefining how banks operate. AI models are now commonly applied to improve operational efficiency, customer service quality, credit scoring accuracy, and fraud prevention in Chinese banks. Across major institutions, the number of AI calls runs into the billions annually, and AI-generated content (code, reports, analyses) has become a routine part of workflows. Banks are moving beyond isolated experiments to enterprise-wide AI platforms – for example, CCB’s model-as-a-service architecture or ICBC’s multi-model matrix – signaling a maturation of AI usage from shiny new tech to fundamental infrastructure.

It’s worth noting that while the big banks forge ahead, smaller lenders face challenges in this AI race. Many regional and mid-size banks lack the massive data sets and R&D budgets to train large models from scratch. They often take a “wait and see” approach, adopting proven solutions a bit later. An interesting development here is that ICBC has started exporting its AI capabilities to smaller banks as a service.

In 2024, ICBC announced it was sharing its large-model (LLM) technology with some small and mid-sized banks, an unusual move in a traditionally siloed industry. This suggests a future where leading banks might provide AI platforms to others (as a utility or service bureau model), helping raise the technology baseline across the sector.

Overall, Chinese banks are entering a new phase of AI-driven transformation – one defined by scale and integration. As one analysis puts it, banks are “embedding models where they generate tangible business value”, not just as gimmicks. The competitive edge is shifting toward those who can harness AI deeply (e.g. ICBC, CCB, BoC), while those slower to invest could fall behind in efficiency and innovation. The Chinese banking industry’s AI evolution is closely watched globally, given its combination of state support, big-data advantages, and now a cohort of homegrown AI models powering critical financial systems.

Industry Trends in AI Adoption

The banking industry globally, and in China specifically, is experiencing a rapid shift toward AI-driven transformation. Key trends include:

China’s Strategic Emphasis on AI: China has prioritized AI as a key driver of economic and technological supremacy. Government policies encourage research and development (R&D) in AI, with financial incentives and regulatory support for innovation, as evidenced by recent government initiatives pushing for locally developed AI platforms like DeepSeek.
Collaborative Initiatives Between Banks and Tech Firms: Major Chinese banks, including ICBC, collaborate with technology firms like Baidu, Tencent, and Huawei to develop AI solutions. These partnerships foster innovation and enable the sharing of insights and resources across the industry, creating a fertile ground for advancements.
Government Support and Policies: The Chinese government actively promotes AI adoption through policies that encourage the use of locally developed AI platforms, such as DeepSeek, and supports R&D with financial incentives. Regulators are also working on comprehensive AI governance frameworks to balance innovation with ethical and security concerns, as seen in recent regulatory developments.
Global Trends in AI Adoption in Banking: Globally, AI is transforming banking operations, from customer service to risk management. Generative AI is particularly impactful, with estimates suggesting it could generate $200 billion to $340 billion in annual value for the banking sector, according to McKinsey, reflecting a worldwide AI fervor driven by technologies like ChatGPT.

Competitor Initiatives

To contextualize ICBC’s efforts, here is a snapshot of what some of ICBC’s major competitors are doing with AI and generative AI:

Bank of China (BoC): Developed a proprietary large model (leveraging DeepSeek R1) for internal use. This model is used for tasks like automated coding, document drafting, and business intelligence. In one month, BoC’s AI produced over 13 million lines of code and served 3,600+ internal users, demonstrating its utility for software development and report generation. BoC also launched an “AI Industry Development Action Plan” pledging heavy investment (reportedly ¥1 trillion credit line) to AI-related initiatives, signaling top-level commitment.
China Construction Bank (CCB): Built an enterprise-grade AI platform integrating both DeepSeek-R1 and other models. As of 2024, CCB had 168 AI use cases deployed across the bank, with 7,000 model instances in production covering areas like investment analysis, front-line customer service, and employee training. Impressively, 16 core banking processes (from credit analysis to knowledge management) now utilize AI at CCB. In 2025, CCB’s CEO confirmed the bank had fully private-deployed a DeepSeek-based large model for internal financial operations.
Postal Savings Bank of China (PSBC): Chose a self-reliant strategy by developing its own large models (the “Youzhi” model suite). PSBC’s AI caters to diverse needs: coding assistants for IT staff, document auditors for compliance, and intelligent trading bots for treasury. In 2024, PSBC reported generating 1.1 million lines of code with AI help (from 5,000+ developer-side “AI helpers”) and achieving a 90% reduction in time for producing certain documents, reflecting huge efficiency gains.
Joint-Stock and Other Banks: Several joint-stock commercial banks are also active. Ping An Bank, backed by the tech-savvy Ping An Group, has invested in AI for personalized wealth management and launched conversational AI services for customers (leveraging Ping An’s broader AI platforms). China Merchants Bank (CMB) has been collaborating with tech firms to build AI models and is integrating AI across office and customer interfaces. Shanghai Pudong Development Bank integrated a large DeepSeek model into its digital assistant to improve customer Q&A and analytical reporting. Bank of Communications (BoCom), another big state bank, is likewise reported to have launched AI model initiatives alongside its peers.

In summary, while ICBC is a leader with its “Smart Ocean” model and expansive use cases, its competitors are not far behind. The big Chinese banks are collectively pushing the frontier of AI in banking – each with their own flavor (BoC focusing on code generation and R&D, CCB on platformization at scale, etc.). This competitive dynamic is spurring faster adoption and innovation, as each institution showcases successful AI applications and others quickly follow suit. For ICBC, the competition provides both motivation to keep innovating and opportunities to collaborate (as seen in its sharing of AI services) to uplift the industry’s tech capabilities as a whole.

Expected Impact of AI at ICBC

The integration of AI and Gen AI is expected to have profound impacts on ICBC’s operations and the broader banking sector:

Enhanced Efficiency: AI-driven automation will reduce operational costs by streamlining processes like loan processing, compliance, and customer service. For example, automating client queries can save millions annually in labor costs, with ICBC reporting over 200 AI-powered use cases logging more than 1 billion AI calls annually. Operational efficiency gains are perhaps the most tangible impact. AI systems are automating high-volume, repetitive tasks and augmenting human workers in complex ones, effectively multiplying productivity. ICBC’s experience quantifies this: its AI models are undertaking work equivalent to tens of thousands of employees’ output. This translates to faster processing times – for example, loan officers can approve credit in a fraction of the time as AI handles data gathering and preliminary risk assessments. AI-assisted coding has accelerated IT development cycles; when a model can crank out millions of lines of code or quickly prototype software, new digital products and features reach the market sooner.
Improved Customer Experience: Personalized services powered by AI and Gen AI will enhance customer satisfaction and loyalty. Tailored product recommendations and 24/7 virtual assistants will strengthen ICBC’s position in retail banking, aligning with its focus on digital transformation.
Strengthened Risk Management: AI’s ability to analyze vast datasets in real-time will improve ICBC’s ability to detect fraud, assess credit risks, and comply with regulations, reducing financial losses and regulatory penalties, as seen in its use of algorithmic credit advisors and intelligent risk detection. Operational efficiency gains are perhaps the most tangible impact. AI systems are automating high-volume, repetitive tasks and augmenting human workers in complex ones, effectively multiplying productivity. ICBC’s experience quantifies this: its AI models are undertaking work equivalent to tens of thousands of employees’ output. This translates to faster processing times – for example, loan officers can approve credit in a fraction of the time as AI handles data gathering and preliminary risk assessments. AI-assisted coding has accelerated IT development cycles; when a model can crank out millions of lines of code or quickly prototype software, new digital products and features reach the market sooner.
New Revenue Streams: AI-enabled innovations, such as AI-driven wealth management and predictive analytics for institutional clients, will create new revenue opportunities, positioning ICBC as a leader in digital finance, with reports highlighting its proprietary "Industrial Smart Ocean" model supporting workloads once handled by 45,000 employees. AI adoption is expected to help reduce costs and enhance revenues for banks. Automation can lower operating costs (less manual processing, fewer errors leading to losses or write-offs), while better customer insights and faster innovation can drive revenue growth (e.g. through improved cross-selling or capturing new tech-savvy customers). There are also new revenue opportunities as some banks consider commercializing AI solutions – for instance, ICBC offering AI services to smaller banks or fintech partners could become a new business line. Banks that master AI may also gain competitive brand value as “technology leaders,” attracting both customers and top tech talent.

On a broader level, the infusion of AI is expected to help Chinese banks handle the scale of the market more effectively. With hundreds of millions of customers and massive transaction volumes in China, AI provides a way to manage this complexity in real-time. It also positions banks to support new initiatives like digital currency, real-time payments, and inclusive finance, since AI can help securely onboard and service huge numbers of users at low marginal cost. In short, AI and generative AI are becoming core to how ICBC and its peers achieve high-quality growth, enabling them to serve the economy with greater agility and insight.

Risks and Challenges

Despite its potential, AI adoption at ICBC and in the banking sector faces several risks and challenges:

Cybersecurity Risks: AI integration introduces new cybersecurity challenges, such as the spread of false information, facilitation of fraud, and cyberattacks. Generative AI-related incidents have increased 53-fold since late 2022, highlighting the need for robust AI-driven threat detection systems, as noted in recent central bank analyses.
Regulatory Compliance and Ethical Concerns: Ensuring compliance with China’s evolving AI regulations and ethical standards is critical. Banks must balance innovation with societal values, avoiding biases in AI models and ensuring transparency in AI-driven decisions, with ICBC investing in governance frameworks to address these issues. Financial regulators in China are closely watching AI developments to ensure stability and fairness. Banks must ensure their AI usage complies with a web of rules: content regulations (to prevent AI from generating disallowed content), algorithmic accountability rules, anti-discrimination laws, and more. For example, China introduced the Interim Measures for Generative AI Services in 2023, which require providers of generative AI to the public to undergo security assessments and abide by content censorship and data security requirements. While these measures mainly target tech firms providing AI chatbots to consumers, banks too must be cautious if any of their AI interfaces interact with the public. Additionally, the financial sector has its own guidelines – e.g. the PBoC and regulators have standards for evaluating AI algorithms in financial applications and guidelines on AI-related information disclosure. ICBC and peers need to build compliance into the design of AI systems (so-called “ethical AI” principles). This means ensuring AI decisions are fair (avoiding bias against protected groups), transparent to regulators, and fall within existing consumer protection laws.
Data privacy and security concerns are paramount. Banks deal with highly sensitive customer information, and training AI models on such data raises the risk of leaks or misuse. Analysts warn that because generative AI systems rely on large datasets, the risk of data leakage increases if not properly controlled. An AI model that memorizes personal data or confidential financial details could inadvertently expose that information. Ensuring compliance with privacy laws (like China’s Personal Information Protection Law) and cybersecurity regulations adds complexity to AI projects. ICBC has responded by strengthening internal governance – the bank’s sustainability reports indicate it has set up an AI security management framework with clear responsibilities, aiming to prevent algorithmic violations and protect data. Nevertheless, maintaining robust data protection while still leveraging data for AI is an ongoing tightrope walk.
Talent Acquisition and Skill Development: The success of AI initiatives hinges on the availability of skilled professionals. ICBC invests in comprehensive training programs and collaborates with academic institutions to address this challenge, recognizing the importance of talent in maintaining a cutting-edge AI workforce. The talent and culture gap is another hurdle. Banks traditionally are not tech companies, and thousands of employees may not be well-versed in AI. ICBC employs over 30,000 technology staff (as of 2024, about 8.6% of its workforce) but scaling AI literacy across the whole organization is difficult. The bank and its peers must train their workforce to understand and appropriately use AI tools. Observers note that banks should “strengthen internal talent training” so staff can keep up with rapid model iterations Change management is essential – persuading veteran bankers to trust and effectively collaborate with AI systems might require cultural shifts and strong support from leadership
Integration with Legacy Systems: Integrating AI with ICBC’s existing legacy systems poses technical challenges. Modernizing infrastructure while maintaining operational continuity requires significant investment and expertise, as part of its digital transformation strategy.
Over-Dependence on AI: Over-reliance on AI may reduce operational resilience, increase the “black box” effect, and suppress diversity of thought, risking groupthink and confirmation bias, as highlighted in central bank reports on AI’s impact on financial stability. Another challenge is model accuracy and reliability. AI systems, especially generative ones, can sometimes produce incorrect or biased results – what’s known as “hallucinations” or errors. In a banking context, an AI error could have serious consequences, such as wrongly denying a loan or flagging a legitimate transaction as fraud. There have been noted concerns that AI mistakes might harm an individual’s credit score or public reputation if not checked. Therefore, ICBC and others must implement thorough validation, human oversight, and feedback loops for their AI models. Achieving a high level of model interpretability is also crucial – regulators and bank risk managers will want to know why an AI made a certain decision. Experts recommend developing tools to visualize and explain AI decision logic, especially for complex LLMs used in credit decisions. ICBC will need to invest in such explainability to satisfy regulatory scrutiny and build trust in AI-driven outcomes.

In summary, the challenges range from technical to organizational to regulatory. ICBC will have to continuously invest in secure infrastructure, rigorous testing, and employee training to mitigate these risks. It will also engage closely with regulators, likely sharing best practices on AI governance as the regulatory environment evolves. Despite these challenges, the consensus is that the benefits of AI – if managed properly – outweigh the risks, which is why ICBC and others are forging ahead carefully rather than pulling back entirely. The onus is on the banks to deploy AI in a responsible, secure, and human-centric manner to fully realize its potential

Regulatory Environment in China

The use of AI in China’s financial industry operates under an increasingly well-defined regulatory framework. Chinese authorities have adopted a stance of encouraging innovation in AI while also imposing safeguards to manage its risks. For banks like ICBC, this means navigating general AI regulations as well as sector-specific supervisory requirements.

A landmark policy is the Interim Measures for the Management of Generative AI Services (often simply called the “AI Measures”), which came into effect on August 15, 2023 These rules – the first of their kind in China – set baseline obligations for any organization providing generative AI services to the public. Key provisions require generative AI service providers to ensure data security, prevent unlawful content generation, label AI-generated content, and protect user privacy. For instance, providers must conduct security assessments and file with the cybersecurity regulator if their AI service could influence public opinion or has societal mobilization capacity. While a bank using AI internally (or for narrow customer interfaces) may not trigger all these rules, the Measures still influence best practices – banks will avoid using public or black-box AI models that aren’t compliant, and will align their AI outputs (e.g. financial advice bots) with permitted content standards (China mandates AI content uphold core socialist values and not produce prohibited material).

Specific to finance, regulators have issued tailored guidelines to ensure AI doesn’t undermine financial stability or consumer protection. The China Banking and Insurance Regulatory Commission (CBIRC) and the central bank (PBoC) have previously emphasized principles for Fintech development, which include responsible use of AI. Notably, there are industry standards like the Evaluation Specification of AI Algorithms in Financial Applications and the Guidance on Information Disclosure for AI in Financial Applications. These likely urge banks to rigorously test AI models for reliability and disclose how they use algorithms in services like robo-advisors or credit scoring, respectively. Additionally, China’s data laws (Cybersecurity Law, Data Security Law, and PIPL for personal data) apply to AI deployments – for example, if ICBC profiles customers using AI, it must respect consent and data minimization rules, and any cross-border data transfer for AI training is tightly regulated by the PBoC’s data security measures

China’s regulatory framework for AI, particularly Gen AI, is among the most comprehensive globally, reflecting the government’s focus on balancing innovation with social and economic stability. Key aspects include:

Content Moderation: Regulations require traceability, authenticity, and watermarks for AI-generated content, with mandatory reporting of illegal content, as outlined in the 2023 Generative AI Measures.
Data Protection: The 2021 Personal Information Protection Law (PIPL) governs data protection, requiring consent for personal data use in AI applications, applying to banking AI systems.
Algorithmic Governance: Security assessments, ethical standards (e.g., no discrimination), and filing with the Cyberspace Administration of China (CAC) are mandatory for AI services, ensuring compliance for customer-facing AI in banking.
Upcoming Legislation: A general AI law is planned, as per the State Council’s 2023 Legislative Work Plan, which will further refine requirements for AI in banking, shaping future compliance needs.

For ICBC, compliance with these regulations is essential, particularly in areas like data protection, algorithmic ethics, and customer-facing AI services, with the bank actively engaging with regulators to contribute to policy formulation.

Importantly, regulators are encouraging innovation alongside caution. The PBoC has publicly supported banks to digitize and use AI for improving financial inclusion and risk control, as long as they “actively prevent and control risks”. We see a coordinated approach: for example, when multiple banks deployed DeepSeek models in early 2025, experts and regulators stressed compliance with relevant laws and supervisory requirements, and advised banks to enhance model interpretability and train staff accordingly. There is recognition that AI can boost efficiency, but also a reminder that banks are accountable for AI-driven decisions. The regulatory environment is thus one of guided usage – encouraging banks like ICBC to lead in AI, but under watchful oversight to ensure stability, fairness, and security are maintained.

Going forward, we can expect even more refined rules. Chinese regulators may introduce certification processes for financial AI models, or audits for algorithms used in critical decisions (similar to stress tests but for AI). As of late 2025, additional national standards for AI (on security, dataset management, etc.) are scheduled to come into effect, which banks will need to align with. Globally, discussions on AI regulation (e.g. the EU’s AI Act) also influence China’s approach. ICBC, given its stature, will likely be a key stakeholder in shaping and complying with these evolving regulations. For now, the bank treads a careful line: aggressively pursuing AI’s benefits, while building compliance into its AI strategy (ensuring, for example, that AI decisions are auditable and can be overridden by humans in sensitive cases). This balance is critical to maintain trust among customers, investors, and regulators as AI’s role in banking expands.

Conclusion

ICBC’s extensive embrace of AI and generative AI reflects a transformative moment for the banking sector. From early chatbots and risk models to today’s large-scale deployment of proprietary LLMs, ICBC has moved swiftly to infuse AI into the fabric of its operations. The bank’s “Industrial Smart Ocean” model and hundreds of AI use cases exemplify how AI can drive efficiency, improve decision-making, and open new service frontiers in finance. At the same time, ICBC’s journey is taking place in parallel with industry-wide advancements – Chinese banks collectively are pushing the envelope, supported by domestic AI innovations like DeepSeek and under the guidance of a proactive regulatory framework. The competitive yet collaborative environment is yielding rapid learning and propagation of best practices across institutions.

The expected impacts – from cost savings and faster innovation cycles to better risk mitigation – are beginning to materialize, positioning banks that lead in AI to outperform in the digital era. However, ICBC’s experience also underscores the importance of managing risks: robust data governance, algorithmic accountability, and human oversight are all vital to safely harness AI’s power. ICBC’s establishment of AI governance systems and its compliance-driven approach show a recognition that trust is as important as technology in financial services. Banks must ensure AI remains a tool to enhance human judgment, not replace it in critical matters without safeguards.

Looking ahead, ICBC and its peers will likely delve deeper into generative AI (perhaps developing even larger multimodal models), fintech partnerships, and cross-sector AI applications. The bank’s forays into sustainable finance and supply chain platforms hint at how AI could help banks diversify services and contribute to societal goals. We can anticipate that AI will become as ubiquitous in banking as core banking systems themselves – an underlying layer that handles myriad tasks instantly and intelligently. For ICBC, staying at the forefront might mean continuously upgrading its AI models, cultivating tech talent, and possibly even helping shape global standards for AI in finance.

In conclusion, ICBC’s use of AI and generative AI showcases both the promise and responsibility that come with cutting-edge technology in banking. The initiatives at ICBC serve as a microcosm of the larger trend: finance is being reinvented by AI, and those who adeptly integrate these technologies – while managing the risks – will lead the industry into the future. ICBC’s path thus far suggests it aims to be one of those leaders, leveraging AI to drive high-quality growth, superior customer service, and innovations that could redefine banking in the years to come. The journey is ongoing, but ICBC’s strides offer valuable lessons in how a legacy bank can successfully transform itself in the age of AI.

Sources:

Industrial and Commercial Bank of China – ESG/FinTech achievements (AI applications, RPA, digital human) icbc-ltd.comicbc-ltd.com
China Innovation Watch – “China's banks are training massive AI models” (ICBC’s 200+ AI use cases, 1 billion calls; Industrial Smart Ocean model replacing 45k employees) ciw.news
Global Times – “Multiple banks deploy DeepSeek AI models...” (ICBC integrating DeepSeek, large-model matrix with 10+ LLMs and 2000 models; 20 business areas and 200 applications with AI; ChatDealing system; credit approval AI assistant; experts on benefits and risks) globaltimes.cnglobaltimes.cnglobaltimes.cn
Yicai Global – “Banking Giants Working on AI…” (At least 11 big banks including ICBC, BoCom, ABC, BOC, CMB launched AI models; integrating AI across office, customer service, research, risk, etc.; AI benefits in efficiency, credit and fraud prevention) yicaiglobal.comyicaiglobal.com
ICBC 2024 Annual Results (ICBC exporting its LLM capabilities to smaller banks; first to achieve 24h automated settlement in custody business) v.icbc.com.cn
The Asian Banker Awards 2023 – ICBC’s FinTech accomplishments (RMB¥10 billion-level AI model applications; AI in investment research and internal audit assistants) icbc-ltd.com
Twimbit Insight on ICBC (Digital transformation initiatives: ICBC e-Credit for supply chain finance across 2000 industrial chains; “ICBC Brains” AML system; voiceprint risk control; smart cooling in data centers; upgraded intelligent service robot “Gong Xiao Zhi”) content.twimbit.comcontent.twimbit.com
Reuters – China Construction Bank deployed DeepSeek R1 model (confirmation of CCB’s internal generative AI deployment in 2025) reuters.com
White & Case – AI Regulatory Tracker (China) (China’s Interim Measures for Generative AI effective Aug 2023; sectoral AI guidelines for financial industry; data/Cybersecurity laws impacting AI use) whitecase.comwhitecase.com
White & Case – AI governance global trends (Risks of AI errors harming credit scores or being misused, highlighting need for regulatory response)whitecase.com