Against the backdrop of global enterprises accelerating the application of artificial intelligence, Asian digital asset financial services group HashKey Group is also systematically advancing the implementation of its AI strategy.
Recently, HashKey Group officially established the "Group Technology Coordination Committee," which is directly overseen by the company's management to plan and execute the overall strategy and implementation path for AI and cutting-edge technologies. According to an internal notice, the committee will be responsible for the top-level design of the group's technical architecture, AI strategic planning, and cross-business line technical coordination, accelerating the organization's transformation toward higher levels of intelligence and automation.
This organizational move also marks a new phase in HashKey's exploration of AI. Previously, AI existed within the company more as an individual tool or in localized trials. As model capabilities improve and organizational awareness matures, HashKey Group has begun systematically advancing AI at the group level and integrating it into internal operations, R&D processes, and user service systems.
HashKey Group CTO Devin Zhang recently shared his observations and insights on the company's AI deployment, security framework, and application prospects in the digital asset industry.
Devin believes that the current moment is a critical juncture for enterprises to widely adopt AI. On one hand, foundational model capabilities have reached a high level, supporting more enterprise-level applications. On the other hand, after years of market education and practical exposure, organizational and talent readiness is gradually improving. In this context, HashKey views AI as an upgrade in organizational capabilities across all processes.
Q: Why is HashKey systematically advancing AI at this stage?
Devin Zhang: Over the past year or two, AI has already been put into practical use within the company, especially in product and R&D teams, where AI-assisted programming is quite common. What has truly changed now is that AI's role is evolving from an individual efficiency tool to a systematically introduced capability at the company and group level. There are two main reasons why this is a critical juncture. First, the foundational capabilities of large models have matured relatively well. Although they are still iterating quickly, the underlying capabilities are already sufficient to support enterprise-level applications. Second, organizational and talent readiness has gradually matured. After years of exposure and use, people have developed a basic understanding and accumulated experience. For HashKey, the goal of advancing AI is to combine human judgment with AI's execution and efficiency capabilities, supporting larger-scale business growth under compliant and controllable conditions.
Q: Where is the most direct value of AI for institutions like HashKey?
Devin Zhang: I think there are two main aspects. One is the systematic improvement of internal operational efficiency, and the other is the continuous enhancement of user experience. For internal operations, the current focus is on two main lines. One is efficiency improvements in the product and R&D chain, meaning the gradual introduction of AI capabilities from requirement breakdown, design, development, testing, to delivery. The other is non-R&D chains, including teams like HR, legal, finance, compliance, marketing, and public relations. Many departments are already using AI, but it's still more of a point-based usage. What's truly important is the AI-ification of the entire chain, which enhances the collaborative efficiency of the entire organization. For the user side, many financial service interactions in the future will shift from operation-driven to intent-driven. Users will express what they want to do, and the system will understand the intent, organize the execution path, and seek user confirmation at key points.
Q: Which AI scenarios is HashKey prioritizing for implementation at this stage?
Devin Zhang: We are more focused on scenarios with clear business value, long time consumption, high repetition, and measurable efficiency improvements. In the product and R&D system, this approach is mainly reflected in the AI-ification of the R&D CI/CD chain. In non-R&D departments, it is more about automating various highly repetitive, rule-based processes.
At the same time, HashKey has also advanced deeper AI applications in infrastructure security and risk control. In infrastructure security, HashKey has already used AI for threat hunting, case tracing, potential risk mining, and IT asset management, helping to enhance the depth and breadth of overall security capabilities. In risk control, for anti-money laundering investigations, group behavior identification, and some complex account issue judgments, an intelligent agent collaboration mechanism has been introduced. Some support processes that previously required the risk control team to spend considerable time handling can now be analyzed by intelligent agents first, followed by review and delivery by relevant personnel. On the R&D side, HashKey has also formed a clear plan for the end-to-end AI-ification of the R&D process. Related explorations are gradually advancing around requirement understanding, code generation, testing, pre-launch audits, and application security audits are also gradually being incorporated into the AI R&D chain.
Q: Why is a security framework a prerequisite for financial institutions to advance AI?
Devin Zhang: Because once AI enters business processes, the object of management has changed. In the past, people focused more on whether the model was strong or its answers were accurate. But when AI agents truly enter enterprise processes, they face system access, interface calls, data reading, process execution, and even triggering external actions. At this point, what enterprises need to manage is an execution entity that can access resources, call permissions, and complete actions. For financial institutions, the focus of risk also falls more on the intelligent agent architecture, permission management, and execution boundaries. To let intelligent agents perform tasks, permissions must be granted to them. Once permissions enter real business flows, permission boundaries, key management, resource invocation rules, behavior tracking, and responsibility attribution must all be redefined. Only after establishing a hierarchical, partitioned, controllable, traceable, and auditable governance system can AI truly enter core business processes.
Q: As a compliant exchange, how will HashKey manage the pace and boundaries of advancing AI?
Devin Zhang: We will systematically advance AI and believe that AI will bring significant positive effects to the industry, company, and business. At the same time, the pace of advancement needs to align with the regulatory environment. At this stage, it is more suitable to prioritize internal efficiency, backend capabilities, risk management, and R&D chain AI-ification, as these areas have clear value, high certainty of efficiency gains, and easier control of external risk exposure. As for user-side innovation, especially capabilities that directly enter the trading process and may significantly change user trading frequency and behavior, the pace of advancement will be more cautious. Intent-driven interaction itself can help simplify user operations, but when intelligent agents further enter automated strategy execution or even replace users in making higher-frequency trading decisions, compliant exchanges need to more fully assess responsibility boundaries, user protection, and regulatory requirements. Whether such capabilities enter the product system is more suitable to be gradually advanced in alignment with the regulatory environment.
At the infrastructure level, HashKey will also adopt a parallel approach using external enterprise-level platforms and private deployment. At this stage, the company will use more platforms with organizational-level data separation, clear security responsibility definitions, and multi-model invocation capabilities. For higher-sensitivity scenarios, a private deployment path will be retained. HashKey itself focuses more on building intelligent agent systems tailored to business needs on top of foundational models.
From a longer-term perspective, HashKey's understanding of AI also extends to the evolution of digital financial infrastructure. Group Chairman and CEO Xiao Feng recently mentioned in a media interview that artificial intelligence and encryption technologies are gradually moving toward deep integration. With the rapid development of AI Agents, intelligent agents may gradually possess independent digital identities and payment capabilities and play more roles in the on-chain economic system. In such a trend, blockchain technology may also become an important infrastructure for managing and coordinating AI Agents.
Devin believes that AI will gradually change the way financial service institutions interact with users, their backend capabilities, and their technical architecture. For HashKey, the short-term goal is to improve efficiency and experience, the mid-term focus is on strengthening backend capabilities and technical foundations, and the long-term goal is to participate in the evolution of a new generation of digital financial infrastructure. For a compliant digital asset institution, this advancement path is more pragmatic.
