Artificial intelligence is rapidly reshaping how retail investors in China and Hong Kong approach the stock market, with major brokerages integrating large language models into their platforms and enabling AI agents to analyze portfolios, generate strategies, and even execute trades. The shift is accelerating, but it is also generating significant regulatory uncertainty as authorities grapple with the risks of AI-driven investment advice at scale.
East Money and the OpenClaw Integration
In April, East Money, China’s largest retail brokerage by active users, organized a virtual trading competition in which participants deployed an OpenClaw AI agent to manage a simulated portfolio, with a prize of 500 yuan for the winner. While the event used a simulated environment rather than real capital, the underlying technology is already live. East Money and other major Chinese brokerages have rolled out functions that allow retail users to connect OpenClaw directly to their account databases, enabling the agent to pull holdings data and generate recommendations.
Chinese firms maintain strict internal bans on employees using AI for professional trading, citing data privacy and execution risk concerns. The asymmetry, AI tools available to retail clients but not to professional staff, reflects the unresolved tension between innovation and fiduciary responsibility that regulators across the region are now being forced to address.
Futu Holdings Leads With DeepSeek-Powered Skills
In Hong Kong, Futu Holdings, which operates the Futubull and Moomoo platforms, has moved more aggressively. In March, Futu launched its AI assistant “Skills,” built on the DeepSeek large language model. Skills allows clients to install the assistant on platforms including OpenClaw, Claude Code, and Cursor, and to execute trading strategies with a trading password as the final confirmation step. Futu has also introduced services that generate quantitative trading strategies from natural language descriptions, lowering the technical barrier for retail investors who want to implement systematic approaches.
Vincent Yao, head of Futu’s AI growth center, noted that the company has observed a positive correlation between AI usage and trading activity: “AI’s impact on trading may be subtle in the near term, but its long-term significance is undeniable.” His colleague Daniel Tse, Futu’s managing director, acknowledged the deeper challenge: “It’s not just about regulatory compliance. It’s also about how you train AI to ensure its recommendations are unbiased. That’s a core challenge the entire industry is still trying to solve.”
Tiger Brokers’ TigerAI product has also seen explosive growth, with cumulative global interactions surpassing 10 million by the end of March, a 500 percent jump since launch one year earlier, and a total user base that grew 148 percent over the same period. Jingxin Du, TigerAI’s product lead, described the competitive stakes: “The next stage of competition among brokerages will hinge on how effectively firms can turn information into intelligence, shifting focus from transaction speed to insight creation.”
Regulators Struggle to Keep Up
The regulatory response has been cautious. The Hong Kong Securities and Futures Commission (SFC) classified AI investment recommendations as “high-risk use cases” in a November 2024 circular. In Singapore, the Monetary Authority (MAS) released an AI Risk Management consultation paper in November and launched a compliance toolkit in March. On the mainland, a Shanghai securities firm was fined 2 million yuan in 2025 for failing to disclose the limitations of its AI-generated investment recommendations.
David Friedland, Asia-Pacific managing director at Interactive Brokers, captured the regulatory dilemma: “AI is changing so fast that the regulation is going to be catching up for a while.” His warning about retail investors is equally pointed: “All of a sudden, with AI, everyone can look like the world’s greatest options trader, but the reality is they need training and experience to understand the risk behind their trading.”
The performance record of AI trading systems provides additional grounds for caution. In a late 2025 competition hosted by Nof1, eight top AI models traded US tech stocks over two weeks; only six of 32 attempts yielded positive results, with a Grok 4.20 model winning with a 34.59% return. The results underscore that AI trading tools are not a reliable path to outperformance, even as they become increasingly accessible to retail investors who may not fully appreciate the underlying risks. This trend reflects a broader shift in the financial sector, where China’s 15th Five-Year Plan has elevated AI to core national infrastructure, accelerating adoption across the entire economy.
The coming months will be a critical test of whether the industry can self-regulate effectively or whether regulators will be forced to intervene more decisively. The SFC’s classification of AI investment recommendations as high-risk use cases suggests that Hong Kong, at least, is prepared to act if the risks materialize. On the mainland, where the retail investor base is larger and the regulatory framework is still evolving, the stakes are correspondingly higher.