Vietnam Proposes Pilot for AI to Place Stock Orders for 1,000 Investors

Vietnam's Ministry of Finance has drafted a regulation that would let artificial intelligence place stock market orders on investors' behalf without human approval for each trade. The program is set to launch as a regulatory sandbox with a hard cap on participants and the exclusion of the market's most important stocks from the index.

How the Pilot Works

The draft establishes a regulatory sandbox for fintech solutions in the securities sector. Under it, technology and brokerage firms would be able to test AI tools that not only offer investment advice but also independently send buy and sell orders to the exchange, without requiring an investor to manually approve each transaction.

The key restriction concerns scale: in the AI order segment, the number of participants cannot exceed one thousand, and only investors qualified as professional. Other fintech solutions tested within the same sandbox have no fixed participant cap - the State Securities Commission will decide the scope of testing case by case.

Why the Largest Stocks Are Excluded

The most liquid and systemically important stocks were excluded from the program from the outset. Shares included in the VNX AllShare index, which covers companies listed on the Ho Chi Minh Stock Exchange and accounts for nearly three-quarters of total market capitalization, cannot be subject to automated AI-driven trading under the pilot.

The measure is meant to limit the potential impact of an algorithm failure, a model error, or coordinated action by multiple AI systems on the prices of the most important stocks and on the index as a whole. The regulator wants to isolate the experiment from the market segment whose destabilization would carry the greatest systemic consequences.

Who Can Participate and How Oversight Works

Brokerages and fund management companies under the regulator's warning or control supervision are excluded from the program, as are entities with accumulated financial losses. All participants will be required to ensure accurate, fair, complete, and timely disclosure of information, without misleading the market, and to comply with anti-money laundering rules.

The State Securities Commission retains the right to suspend or fully halt the tests at any time if a threat emerges to the security of the financial market, IT infrastructure, public order, or the general interest. This emergency mechanism is meant to give the regulator a way to react quickly before a single AI system's error translates into broader market disruption.

Regional Context and Risks

The proposal comes amid a growing number of individual investors in Vietnam using AI tools for analysis and investment decisions, a trend local media have previously described as raising concerns about liability for losses incurred based on an algorithm's recommendations. Vietnam's securities regulator has also previously examined the risks of high-frequency algorithmic trading, pointing to possible market destabilization.

The regulatory sandbox is meant to answer that tension: on one hand, allowing the development of technology that is genuinely changing how retail and professional players invest, and on the other, limiting the scale of any potential harm to the test group and a select set of stocks outside the main index.

What This Means for the Polish Market

For Polish financial institutions and regulators, this is another example of how regulators worldwide are trying to admit autonomous AI systems into capital markets without waiting for comprehensive, general legislation. The test model, with its participant cap and exclusion of the most important stocks, could serve as a reference point for discussions on how to admit similar solutions in the European Union, where oversight of AI in financial services is still taking shape.

For technology companies developing algorithmic trading tools, the Vietnamese draft also shows that regulators expect not just model effectiveness but, above all, emergency shutdown mechanisms and full operational transparency toward supervisors before a system gets access to real orders on a live market.