ICBC, with RMB57 trillion of total assets, runs its AI on an in-house large-model platform, ICBC Zhiyong, which it says has hundreds of billions of parameters and was deployed in more than 600 scenarios by mid-2026; its 'AI-ICBC' plan adds an agent factory, two super agents and an AI workload it measures in tens of thousands of person-years.
How is ICBC using AI?
Industrial and Commercial Bank of China (ICBC), which reported RMB57,070,593 million in total assets at 30 June 2026, builds its AI on ICBC Zhiyong, an in-house financial large language model system with hundreds of billions of parameters. Its own reports chart the scale-up: more than 200 scenarios at the end of 2024, more than 500 in over 30 business areas at the end of 2025, and more than 600 by 30 June 2026. ICBC measures the effect as workload: AI digital employees handled the work of 55 thousand persons a year in 2025, up from more than 40 thousand in 2024. In 2026 it upgraded its D-ICBC strategy to AI-ICBC, built two super agents (one for customers, one for employees), put a risk-control agent into production and moved four categories of front-line agents into pilot. It invested RMB28,588 million in FinTech in 2025. President Liu Jun said in October 2026 that the bank carries the liability for its AI, and described a 'proof of benefit' accountability framework.
Lead bank
Industrial and Commercial Bank of China Limited
Total assets (group, as reported, Jun 30, 2026)
RMB57,070,593 million total assets[1]
Charter
State-controlled joint stock commercial bank incorporated in the People's Republic of China, listed in Shanghai (A shares) and Hong Kong (H shares), supervised by the National Financial Regulatory Administration (NFRA), with overseas subsidiaries including ICBC (Asia) in Hong Kong
Headquarters
Beijing, China
Flagship platform
ICBC Zhiyong and the AI-ICBC '1+1+3' system
AI leadership
Liao Lin · Liu Jun · Zhao Guide · Wang Jingwu · Tian Fenglin · Tony Mak Wing-yin
- Platform: ICBC Zhiyong, an enterprise-level financial LLM system with over 100 billion parameters, built in-house; the 2024 annual results say it had been called more than 1 billion times and won the People's Bank of China's first prize for FinTech development.
- Scale: LLM scenarios grew from more than 200 in over 20 business fields (2024) to more than 500 in over 30 (2025) and more than 600 (first half of 2026), according to ICBC's annual and interim results.
- Workload measure: ICBC reports AI and digital employees doing the work of more than 30 thousand persons (2023), more than 40 thousand (2024) and 55 thousand persons a year (2025). It does not publish a revenue or cost figure for AI.
- Strategy: in 2026 D-ICBC became AI-ICBC, organised as '1+1+3' (the Zhiyong technology foundation, the ICBC D-Space data space, and three platforms for corporate customers, personal customers and employees), the model Senior Executive Vice President Zhao Guide set out on 28 August 2026.
- Agents: two super agents (customer-facing GINO and an employee agent), four categories of front-line agents in pilot, and an 'Intelligent Risk Control' agent in production by mid-2026; the customer-manager agent delivered more than 22 million services in the first half of 2026.
- Measured results disclosed on 28 August 2026: credit review assistant supported approvals involving more than RMB1 trillion; 360 thousand compliance review recommendations; a private-banking agent cut solution preparation from several days to three hours; intelligent quote-based trading above 96%.
- Spend and people: RMB28,518 million invested in FinTech in 2024 (3.63% of operating income, 36 thousand FinTech staff, 8.6% of employees) and RMB28,588 million in 2025; more than 12 thousand data analysts across the group by mid-2025.
- Governance: a Digital Finance and Financial Technology Committee reviews major technology matters; the 2025 sustainability report describes algorithm security self-assessment and monitoring across development, use and maintenance, and ICBC says it was first in the industry to pass CAICT's large-model data security capability assessment (2025).
- Risk platform: the enterprise-level intelligent risk management platform covered nearly 300 decision scenarios at the end of 2025 and 323 business scenarios by August 2026, across domestic branches, subsidiaries and overseas institutions.
- Regulation: the NFRA's AI guidelines for banks and insurers (NFRA [2026] No. 8, issued 18 June 2026, as reported by The Asian Banker) require boards to set dedicated AI governance arrangements and institution-wide AI strategies; The Asian Banker names ICBC among the Chinese banks already well advanced.
- Accountability: at the Shanghai Bund Summit in October 2026 Liu Jun said ICBC had deployed hundreds of AI models with more autonomy in data-heavy work, kept humans in client-facing roles, and introduced a 'proof of benefit' framework tying accountability to those who reap the returns (South China Morning Post).
What is ICBC Zhiyong and the AI-ICBC '1+1+3' system?
ICBC Zhiyong is the bank's own enterprise-level financial large language model system, described in the 2024 annual results as having over 100 billion parameters. In 2026 ICBC renamed its digital strategy from D-ICBC to AI-ICBC and organised it as '1+1+3': the Zhiyong technology foundation (computing power, large models, an AI agent factory and an AI risk control system), the 'ICBC D-Space' enterprise data space, and three platforms for corporate customers, personal customers and employees. On top sit two super agents, one facing customers ('GINO', Gong Xiao Zhi, across mobile banking, outlets and the contact centre) and one facing employees, plus agents for customer managers, corporate marketing, credit review and risk control. The 2026 interim results say inference runs with a multi-level cache acceleration solution on domestically developed computing power. ICBC's English reports opened for this page do not name an outside model provider.[7][1][2][3]
Who does ICBC work with on AI?
ICBC's English reports describe ICBC Zhiyong as built in-house and name no outside model provider. The partners they do name are public bodies: ICBC launched a financial-sector privacy computing platform with the National FinTech Risk Monitoring Center, passed CAICT's large-model data security assessment, and its Zhiyong work won the People's Bank of China's FinTech development first prize. The 2026 interim results say ICBC ranked first among peers in patent publications and cumulative patent grants, without an AI breakdown.[7][5][1]
| Item | What it is | Status | Date |
|---|---|---|---|
| National FinTech Risk Monitoring Center | Joint financial-sector privacy computing platform | Launched | 2025-03-28[7] |
| CAICT large-model data security assessment | ICBC first in the industry to pass | Passed | 2025-08-29[5] |
| People's Bank of China FinTech Development Award | First prize for the Zhiyong large-model project | Awarded | 2025-03-28[7] |
What has ICBC done on AI, and when?
- Nov 5, 2026
ICBC urges e-CNY local wallets for foreign firms[17]
At Hong Kong Fintech Week an ICBC executive, Zhang Yi, encouraged overseas institutions to open e-CNY enterprise local wallets; the South China Morning Post reported ICBC is a core e-CNY operator with cumulative volumes in the hundreds of billions of yuan and among the first direct participants on the upgraded CBETS platform.
- Oct 20, 2026
Liu Jun on AI liability and 'proof of benefit'[18]
Speaking at the Shanghai Bund Summit, the president said commercial banks carry direct financial and legal liability for AI, cited four mismatches between AI and banking (disclosure, data storage, sovereignty and credibility), and said ICBC uses a 'proof of benefit' framework tying accountability to those who reap the financial returns.
- Aug 31, 2026
Big six state banks report shared growth[14]
The South China Morning Post reported that ICBC, CCB, ABC, BOC, Bank of Communications and PSBC all grew first-half revenue and profit for the first time since 2022; ICBC's net interest margin edged up to 1.29%, and Liu Jun credited more appropriate liability costs rather than repricing alone.
- Aug 28, 2026
Risk platform reaches 323 scenarios[2]
Wang Jingwu said the enterprise-level intelligent risk management platform had been rolled out to all domestic branches, integrated subsidiaries and overseas institutions, covering 323 business scenarios, and that an AI agent for risk management now serves the first line of defense, risk officers and risk managers.
- Aug 28, 2026
AI-ICBC '1+1+3' set out at the interim results release[2]
Senior Executive Vice President Zhao Guide said ICBC had upgraded D-ICBC to AI-ICBC and was building an AI agent factory; he reported 22 million services by the customer-manager assistant, credit review support for more than RMB1 trillion of approvals, 360 thousand compliance review recommendations, and a private-banking agent that cut solution preparation from several days to three hours.
- Aug 28, 2026
2026 interim: 600 LLM scenarios, two super agents, risk-control agent in production[1]
LLMs were deployed in more than 600 scenarios; ICBC formed a hierarchical framework for agent development, completed two super agents for customers and employees, moved four categories of front-line agents into pilot, put the 'Intelligent Risk Control' agent into production and implemented a multi-level cache inference acceleration solution on domestically developed computing power.
- Mar 27, 2026
Sustainability report sets out AI security management[4]
The 2025 sustainability report says ICBC carries out algorithm security self-assessment, information and algorithm security monitoring, incident response and handling of algorithm violations across algorithm development, application and maintenance, under a FinTech ethics governance system.
- Mar 27, 2026
2025 results: 500 AI applications and 55 thousand persons of workload[3]
ICBC reported over 500 AI applications across more than 30 business areas, AI digital employees handling the workloads of 55 thousand persons a year, a Zhiyong-based agent system of 'one super agent and multiple specialized agents', its first marketing AI agent 'Gong Xiao Cai', the 'Bills Insight 2.0' AI contract review system, and RMB28,588 million invested in FinTech.
- Jan 26, 2026
ICBC (Asia) at Hong Kong's AI for Digital Finance forum[16]
On a trade-finance panel with HSBC and Hang Seng Bank about using AI and shared data, ICBC (Asia)'s managing executive officer Jimmy Jim said geopolitical tensions were accelerating trade diversification while the industry faced intense competition and fragmented platforms.
- Aug 29, 2025
Results release: 96% intelligent trading ratio, AI wealth assistant at 91% satisfaction[6]
ICBC told investors the intelligent ratio of smart inquiry and trading in financial markets had reached 96%, AI supported credit operations for more than 20,000 credit practitioners, the AI wealth assistant had a 91% user satisfaction rate, and an intelligent marketing agent for corporate banking had launched.
- Aug 29, 2025
2025 interim: 100 new AI scenarios and a large-model security assessment[5]
Under the 'AI+' Action Plan ICBC launched over 100 new scenarios in personal banking, financial markets and corporate credit, including an AI wealth assistant and an intelligent investment research assistant, and said it was first in the industry to pass the CAICT 'Trusted Intelligence and Security: Large Model Data Security Capability Assessment'. Data analysts across the group exceeded 12 thousand.
- Mar 28, 2025
2024 results: ICBC Zhiyong in 200 scenarios, 40 thousand persons of workload[7][8]
ICBC Zhiyong, an enterprise-level financial LLM system with over 100 billion parameters, served more than 20 business fields and over 200 scenarios and was called more than 1 billion times; AI and other new technologies did work equivalent to over 40 thousand persons; AI coding assistants were launched for developers, and ICBC exported LLM capabilities to small and medium-sized banks. FinTech investment was RMB28,518 million with 36 thousand FinTech staff.
- May 22, 2024
Liu Jun appointed President[3]
The board appointed Liu Jun as President and elected him Vice Chairman; he became an executive director at the annual general meeting on 28 June 2024.
- Mar 28, 2024
Chairman: RMB80 billion on technology in three years[10]
At the 2023 results briefing Liao Lin said ICBC had spent RMB80 billion on technology over the past three years, called it the first bank to develop a 100 billion-level generative AI model, and said nearly 30 thousand engineers were working on the D-ICBC drive.
- Mar 27, 2024
2023 results: large-model system in operation, digital employees at 30 thousand persons[9]
The 2023 annual results say ICBC was the first among domestic peers to build a billion-level AI big model technology system and put it into operation, and that digital employees took on the workload of more than 30 thousand natural persons.
- Feb 19, 2024
Liao Lin confirmed as chairman[11]
ICBC announced that the NFRA had approved Liao Lin's qualification as Chairman of the Board and Chairman of the Strategy Committee, to which the board elected him on 1 February 2024.
- Feb 15, 2024
ICBC Asia doubles down on AI against scams[12]
The head of financial crime compliance at ICBC (Asia), the group's Hong Kong unit, told the South China Morning Post it would adopt more machine learning and other AI technologies to identify potentially fraudulent transactions and investigate alleged scams, after reported scam cases in Hong Kong rose 40% in 2023.
Where does ICBC use AI today?
| System | What it does | Status | Use case |
|---|---|---|---|
| ICBC Zhiyong large-model platform[7][1][2] | In-house financial LLM system with over 100 billion parameters; more than 600 deployed scenarios by mid-2026; ICBC says it also extends the platform to industry partners and smaller banks. | In production | Generative & agentic AI |
| GINO (Gong Xiao Zhi) customer-facing agent[3][1][2] | Unified customer agent across mobile banking, outlets and the contact centre, delivering a 'Conversation as a Service' model; ICBC also piloted what it calls the first in-branch LLM customer service in China's financial sector. | Rolling out | Customer-facing chatbots |
| Gong Xiao Cai and the customer-manager agent[3][2] | Gong Xiao Cai, launched in 2025, is ICBC's first marketing AI agent for retail client managers; in the first half of 2026 the dedicated agent for personal customer managers provided more than 22 million services. | In production | Generative & agentic AI |
| Credit review assistant[2][6] | Supported review and approval involving more than RMB1 trillion in the first half of 2026; AI supported credit operations for more than 20,000 credit practitioners by mid-2025. | In production | Credit scoring & underwriting |
| Intelligent Risk Control agent and enterprise risk platform[1][2] | Agent providing risk knowledge Q&A, customer health checks and risk data analysis, on a platform covering 323 business scenarios. | In production | Model risk management |
| Intelligent compliance review assistant[2] | Generated 360 thousand compliance review recommendations in the first half of 2026; review comments by compliance staff rose 130% with system assistance. | In production | AI governance (general) |
| Global intelligent conversational trading system[2][6] | Intelligent closed loop across the transaction chain; intelligent quote-based transactions exceed 96%. | In production | Trading & capital markets |
| AI wealth assistant and private-banking agent[5][6][2] | AI wealth assistant launched in 2025 with a 91% user satisfaction rate; a comprehensive financial service agent in private banking cut solution preparation from several days to three hours. | In production | Generative & agentic AI |
| AI coding assistants and intelligent R&D platform[7][3] | AI coding assistants launched for frontline developers in 2024; an enterprise-level intelligent R&D platform covers the whole development process. | In production | AI-generated code & coding agents |
| Online risk control and AI masks[3] | Intelligent risk control on online platforms, with technologies including AI masks against unwanted facial recognition, device fingerprints and remote audio-visual checks to block risky transactions. | In production | Fraud detection |
| ICBC (Asia) machine learning for scam detection[12] | The Hong Kong unit said in February 2024 it would use more machine learning and other AI technologies to identify potentially fraudulent transactions and investigate scams. | Announced | Fraud detection |
| Large-model data security and algorithm security management[5][4] | First in the industry to pass CAICT's large-model data security capability assessment (2025); algorithm security self-assessment and monitoring across the algorithm lifecycle. | In production | Data & privacy |
Use-case links open the tracker's rules-by-use-case matrix: which documents govern each system type.
What has ICBC disclosed in numbers?
| Metric | Value | As of |
|---|---|---|
| Total assets | RMB57,070,593 million[1] | Jun 30, 2026 |
| Large-model scenarios | More than 600[1] | Jun 30, 2026 |
| AI applications and business areas | Over 500 across more than 30 business areas[3] | Dec 31, 2025 |
| Workload handled by AI digital employees | 55 thousand persons a year (2025); over 40 thousand persons (2024); over 30 thousand (2023)[3][7][9] | Dec 31, 2025 |
| FinTech investment | RMB28,588 million (2025); RMB28,518 million (2024)[3][7] | Dec 31, 2025 |
| FinTech staff | 36 thousand, 8.6% of employees; FinTech spend 3.63% of operating income[7] | Dec 31, 2024 |
| Customer-manager agent services | More than 22 million in the first half of 2026[2] | Jun 30, 2026 |
| Credit approvals supported by the credit review assistant | More than RMB1 trillion in the first half of 2026[2] | Jun 30, 2026 |
| Compliance review recommendations generated | 360 thousand; compliance staff review comments up 130%[2] | Jun 30, 2026 |
| Intelligent quote-based trading | Above 96%[2] | Aug 28, 2026 |
| Intelligent risk platform coverage | 323 business scenarios[2] | Aug 28, 2026 |
| ICBC Zhiyong calls | More than 1 billion cumulatively[7] | Dec 31, 2024 |
| Data analysts | More than 12 thousand across the group[5] | Jun 30, 2025 |
| Mobile banking app score (Sia 2025 Mobile Banking Benchmark) | 68, third in Asia behind Ping An Pocket Bank (72) and China Merchants Bank (69)[15] | Dec 7, 2025 |
| Net interest margin, first half 2026 | 1.29%[14] | Jun 30, 2026 |
| Technology spending, three years to 2023 | RMB80 billion[10] | Mar 28, 2024 |
“If something bad happens, I’m going to have to pay the victim and give the final answer to the market.”[18]
Liu Jun · President · Oct 20, 2026
“ICBC’s full-stack, independently controllable large-model platform, “ICBC Zhiyong”, has not only been deployed in more than 600 scenarios, but has also extended to industry partners through an open architecture and shared model, helping small and medium-sized enterprises use artificial intelligence at a lower threshold.”[2]
Liu Jun · President · Aug 28, 2026
“In response to inherent defects of large models and risks from external attacks, we strengthen multi-layered defense in depth to ensure that artificial intelligence is secure, reliable and controllable.”[2]
Zhao Guide · Senior Executive Vice President · Aug 28, 2026
“We launched an AI agent for risk management and control, providing the first line of defense, risk officers and risk managers with digital and intelligent services such as risk knowledge Q&A, intelligent customer health checks and risk control data analysis, effectively reducing the workload at the primary level.”[2]
Wang Jingwu · Senior Executive Vice President · Aug 28, 2026
“We are the first bank to develop a 100 billion-level generative AI model, finding the new quality productive forces taking shape at ICBC after a faster pace.”[10]
Liao Lin · Chairman · Mar 28, 2024
Who runs AI at ICBC?
Liao Lin
Chairman of the Board of Directors and Chairman of the Strategy Committee (elected 1 February 2024)[11][1]
Zhao Guide
Senior Executive Vice President (answered for FinTech and AI-ICBC at the 2026 interim results release)[2][1]
Wang Jingwu
Executive Director, Senior Executive Vice President and Chief Risk Officer (oversees the intelligent risk management platform and risk-control AI agent)[3][2]
Tian Fenglin
Board Secretary (presented the 2026 interim results)[2]
Tony Mak Wing-yin
Head of the financial crime compliance department, ICBC (Asia) (as named in February 2024)[12]
Which regulators govern ICBC's AI?
| Authority | Why it matters here | Documents |
|---|---|---|
| China (NFRA / PBoC / CAC) | ICBC's home supervisors. The NFRA approves its directors and senior managers and its 2024 data security measures require unified management and pre-use review of AI models; the PBC's FinTech plan and technology-ethics guidelines are the texts ICBC's sustainability report cites for its AI governance; and CAC rules govern public-facing generative AI services such as GINO. The Asian Banker reports that on 18 June 2026 the NFRA issued Guidelines on the Safe Development and Application of Artificial Intelligence in the Banking and Insurance Industry (NFRA [2026] No. 8), 32 principles that require boards to set up dedicated AI governance arrangements and institution-wide AI strategies. | NFRA Data Security Measures (金规〔2024〕24号) CAC Generative AI Interim Measures (Order No. 15) |
| HKMA | ICBC is listed in Hong Kong and runs ICBC (Asia) there; its president shared the October 2026 Bund Summit panel on AI accountability with the HKMA's deputy chief executive, whose agency had just updated its AI governance framework; ICBC (Asia) told the South China Morning Post in February 2024 that it was adding machine learning to scam detection. | |
| Basel Committee | ICBC's bank-wide unified indicator system (over 70 thousand indicators) and enterprise data space underpin its AI models, which engages the Basel Committee's BCBS 239 principles for risk data aggregation. | BCBS 239 |
What does the public record suggest about ICBC's AI strategy?
Analysis by BankingNewsAI from the sources cited on this page. Observations, not advice.
01
Workload, not money, is the AI metric
ICBC reports AI output as person-equivalents (55 thousand persons a year in 2025) and scenario counts, not as revenue or cost saved. Comparing it with DBS or JPMorgan on value requires care: its disclosed figures are throughput measures.
02
A sovereign stack by design
The Zhiyong platform is described as full-stack and independently controllable, with inference on domestically developed computing power, and the English reports name no outside model vendor. That makes ICBC a test of how far a bank can scale AI on its own models and chips.
03
Agents move into risk and credit
The 2026 disclosures put agents into credit review (RMB1 trillion of approvals supported) and risk control (an agent in production), not only service and marketing. Those are the areas where the NFRA's model-review rules bite hardest.
Which leader rankings does ICBC appear in?
6 of the computed leaders rankings include ICBC, each ranked from the same sourced records (bank pages, filings, hiring snapshots, the agent tracker, the incidents register).
How does ICBC compare with its peers?
- How does Industrial and Commercial Bank of China compare with its peers on AI and technology?
- Industrial and Commercial Bank of China competitors and alternatives, by business line
Its peer group on this site, by business model and size:
What is ICBC Zhiyong?
ICBC Zhiyong is Industrial and Commercial Bank of China's in-house enterprise-level financial large language model system, described in its 2024 annual results as having over 100 billion parameters. It served more than 200 scenarios at the end of 2024 and more than 600 by 30 June 2026, and ICBC says it has extended the platform to industry partners and smaller banks.
How does ICBC use AI?
ICBC uses AI in customer service (the GINO agent across mobile banking, outlets and the contact centre), customer-manager support (the Gong Xiao Cai agent), credit review, risk control, compliance review, financial-markets trading, wealth management and software development. Its 2025 results count over 500 AI applications in more than 30 business areas.
What is AI-ICBC?
AI-ICBC is the 2026 upgrade of ICBC's D-ICBC digital strategy. It is organised as '1+1+3': the ICBC Zhiyong technology foundation, the ICBC D-Space data space, and three digital and intelligent platforms for corporate customers, personal customers and employees, aiming at 'one customer, one advisor' and 'one position, one assistant'.
How much work does AI do at ICBC?
ICBC reports that AI digital employees handled the workloads of 55 thousand persons a year in 2025, up from more than 40 thousand in 2024 and more than 30 thousand in 2023. It does not publish an AI revenue or cost-saving figure.
How much does ICBC spend on technology?
ICBC invested RMB28,588 million in FinTech in 2025 and RMB28,518 million in 2024, when FinTech spending equalled 3.63% of operating income and 36 thousand FinTech staff made up 8.6% of employees. It does not report a separate AI budget.
Who leads AI at ICBC?
Chairman Liao Lin (since February 2024) and President Liu Jun (since May 2024) front the AI-ICBC strategy. Senior Executive Vice President Zhao Guide answered for FinTech and AI at the August 2026 results release, and Senior Executive Vice President and Chief Risk Officer Wang Jingwu oversees the intelligent risk platform and its AI agent. ICBC's English reports name no chief AI officer.
- 2026 Interim Results Announcement
- Q&A Record for the 2026 Interim Results Release of Industrial and Commercial Bank of China
- 2025 Annual Results Announcement
- 2025 Sustainability Report
- 2025 Interim Results Announcement
- Q&A Record for 2025 Interim Results Announcement of Industrial and Commercial Bank of China
- 2024 Annual Results Announcement
- Q&A Record for Results Presentation of Industrial and Commercial Bank of China in 2024
- 2023 Annual Results Announcement
- Minutes of ICBC 2023 Annual Results Briefing: Investor and Analyst Session
- Announcement on the Appointment of Chairman of the Board of Directors
- ICBC Asia to combat fraud by doubling down on AI amid rise in scams, anti-financial crimes chief says
- Can China's banks turn AI governance from a board mandate into a competitive advantage?
- China's 6 biggest state banks post rare shared growth as margins show tentative rebound
- Mainland China's bank apps the best in Asia, beating Hong Kong and Singapore, study says
- Digital finance forum: AI, data must link 'digital islands' to smooth trade finance
- Chinese banks explore local e-CNY wallets for foreign firms via upgraded platform
- Commercial lenders shouldering heavier AI burden than central banks: ICBC president
When ICBC moves on AI, you'll read it here first.
when one of the banks we cover moves on AI, it is in the next morning's brief · six sourced stories · 7 am PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning