# Ứng dụng AI trong chứng khoán: Dễ mua công nghệ, khó xây niềm tin

- Nguồn: VnExpress Khoa học Công nghệ
- Thời gian phát hành: 2026-08-21 23:00 (giờ Việt Nam)
- Điểm AI: 85/100
- Link AIHOT.vn: https://aihot.vn/items/89c464793ec7fbea
- Link gốc: https://vnexpress.net/ai-trong-chung-khoan-de-mua-cong-nghe-kho-dat-niem-tin-5111879.html

## Tóm tắt AI

Dù việc triển khai các tác nhân AI vào hạ tầng chứng khoán hiện nay khá thuận lợi, nhưng bài toán về độ tin cậy và tính minh bạch vẫn là rào cản lớn đối với các nhà đầu tư và tổ chức tài chính.

## Thân bài

![AI trong chứng khoán 'dễ mua công nghệ, khó đặt niềm tin'](https://i2-vnexpress.vnecdn.net/2026/08/21/dsc04652-jpg-1787327142-2049-1787327289.jpg?w=1200&h=675&q=100&dpr=1&fit=crop&s=ShIqExckTPEMGvXio-3mew)

### AI in the stock market: 'Easy to buy technology, hard to place trust'

Experts assess that purchasing and deploying AI agents within stock market infrastructure is not difficult, but placing trust in them remains a significant challenge.

At the Vietnam Securities Tech Summit 2026 held on August 21 in Ho Chi Minh City, most experts agreed that in the context of the AI boom, digital transformation in the securities sector is almost mandatory to enhance competitiveness.

"Purchasing technology for application is currently very simple, including AI. Most large enterprises now have the financial capacity to acquire the latest solutions," said Mr. Dang Trung Thanh, Chief Information Security Officer at Techcom Securities (TCBS), part of the Techcombank ecosystem.

![Các diễn giả thảo luận tại Vietnam Securities Tech Summit 2026. Ảnh: Bảo Lâm](https://i1-vnexpress.vnecdn.net/2026/08/21/wm1-dsc04652-jpg-1787327103-17-1372-8637-1787327289.jpg?w=1020&h=0&q=100&dpr=1&fit=crop&s=ajMbwpSWokC982lP9v-Dgw)

Speakers discussing at the Vietnam Securities Tech Summit 2026. Photo: Bao Lam

Mr. Mai Tat Thang, Chief Information Officer of VPS Securities, noted that one should not focus too much on buying AI models but rather on upgrading data infrastructure, standardizing processes, and implementing APIs. "Buying an AI model might take only a week, but building a standardized, trustworthy data foundation for AI to function effectively will take 2-3 years," Mr. Thang said.

The development of AI, cloud computing, and Big Data opens up opportunities but also poses challenges regarding cybersecurity, data privacy, and risk management. "The stock market has one very distinct feature: trust," observed Mr. Tran Minh Quang, Director of the Cyber Security Monitoring and Response Center at Viettel Cyber Security. "With the support of AI, the concern is not only about the trading floor infrastructure being attacked, but also whether our trust itself is being attacked."

According to Mr. Quang, with the "deep" support of generative AI, drafting malicious content and fake news online is very easy. These can deceive investors by combining various techniques to manipulate the market, causing a dual impact: affecting information security, data, and privacy, while directly undermining investor confidence. For securities organizations, the consequences can be much more severe.

The Viettel Cyber Security expert assessed that AI has provided organizations, businesses, and individuals with new technological capabilities—a new "arm." However, conversely, their adversaries also possess similar "arms." AI has "lowered the barrier" for attack solutions significantly. Malicious actors do not need deep expertise in cybersecurity to leverage existing technology combined with AI support to launch attacks on securities infrastructure.

"AI is causing the risk of attacks to increase exponentially, not just linearly," Mr. Quang emphasized. However, he also suggested that one should not be overly pessimistic or worried, as AI itself also helps businesses and organizations keep pace with the development of those threats.

Based on corporate practice, Mr. Nguyen Xuan Quang, a representative from SSI Securities Corporation, argued that risks should be assessed by classifying them when deploying AI, but in a flexible rather than rigid manner. The core principle is to prevent clear-cut malicious actions while monitoring and evaluating ambiguous ones to detect attacks as early as possible.

According to Mr. Thang from VPS, there are three factors that need to be controlled regarding AI-executed tasks. First, AI agents must be identified within the system, must not share accounts with human operators, and must be subject to immediate revocation upon request.

Regarding infrastructure, there needs to be a "ceiling" limit, meaning AI operations must be controlled and limited in terms of frequency and risk appetite, with the ability to press a button to stop everything "immediately" if issues arise. Additionally, AI agents should be assigned "accountability," requiring them to report the history of data used, the principles applied, and who approved the actions.

Ms. Dang Thi Thanh Huong, Deputy CEO of the Ho Chi Minh City Stock Exchange (HOSE), proposed that the securities sector in Vietnam must have an ecosystem and data that is "accurate, sufficient, clean, and live." HOSE is currently connecting, sharing, and providing data to securities companies, listed companies, and news agencies, while coordinating with the Department of Information Technology and Digital Transformation of the Ministry of Finance to connect to the national financial database as well as data on issuance and securities trading on both exchanges and national reserves.

"I hope we will soon have an ecosystem that connects all market participants, so that data flows can move and everyone can benefit," Ms. Huong said.

Bao Lam