Securities AI is moving from pilot to production in Vietnam’s financial sector, but the accountability framework has not kept pace. At the Vietnam Securities Tech Summit 2026, held in Ho Chi Minh City on August 21 and organized by IEC Group, experts and regulators described AI, big data, and cloud computing being embedded into market analysis, broker support, and client personalization. The unresolved question they raised: when AI-generated investment recommendations go wrong, who bears the responsibility?
What is securities AI actually doing inside Vietnamese firms?
According to Nguyen Xuan Quang, Assistant Chairman of SSI Securities, AI is being applied across three main areas: real-time customer behavior analysis, personalized investment suggestions, and chatbot or virtual assistant deployment for near-instant query handling. The level of integration goes beyond analytics dashboards into live, client-facing functions.
SSI has drawn a clear internal line. Quang stated at the summit that AI at SSI produces recommendations but does not make final decisions. “AI makes recommendations but does not decide on behalf of humans. Humans must retain the final decision.” This distinction is one few Vietnamese firms have formalized publicly.
Who is liable when AI advice leads to investor losses?
This is the accountability gap the summit surfaced. Quang noted that the risk is not hypothetical: both securities firms and individual brokers are already using AI to analyze stocks and issue recommendations. If an AI recommendation is wrong and an investor loses money, current frameworks do not clearly assign responsibility.
A second layer of risk sits alongside it. AI can generate inaccurate information presented in a convincing way, and it can be used to create fraudulent investment recommendations or spread misinformation, either through impersonation of established firms or by manipulating sentiment on social media platforms.
Why are cybersecurity experts more worried about trust than systems?
Tran Minh Quang, Director of the Cyber Threat Analysis and Sharing Center at Viettel Cyber Security, described a shift in the threat landscape. Previously, launching a meaningful attack on a financial system required significant technical expertise. AI has changed that calculus: actors without deep technical backgrounds can now organize attacks on digital infrastructure.
For securities markets, the more serious risk is not data breach or system downtime. “The biggest risk is not attacking the system but attacking market trust,” Quang said. AI-generated fake content, coordinated misinformation on social networks, and synthetic recommendations can move investor sentiment and affect market behavior without ever directly touching a trading system.
The same tools create defensive value. With sufficient data quality, AI can detect anomalous patterns early, flag potential fraud, and support incident response, reducing pressure on security teams. Quang’s position: AI and human analysts are a pair requiring parallel investment in both technology and personnel. “Even when AI is automated, there must still be humans in control and humans who are accountable.”
Is AI governance keeping up with AI adoption?
Tuan Vo, Senior Solutions Architect at G-AsiaPacific Vietnam (GAPV), identified a gap common across the sector: staff at many securities firms are already using AI tools independently, without a centralized governance framework. The consequences are twofold. Infrastructure and model costs become difficult to track and control. And fragmented AI use creates risk around sensitive financial data.
“Securities firms do not just need separate AI tools. They need a unified AI platform that controls costs, protects data, and optimizes performance,” Vo said at the summit. GAPV presented an Agentic AI (AI that can take multi-step actions autonomously, not just answer single questions) approach built on AWS Cloud, combining Amazon Quick for data connectivity across internal systems, cloud sources, and SaaS applications, with role-based access control, security settings, monitoring, and audit logging. The stack also includes Kiro IDE, a developer environment that integrates AI across the full software development lifecycle from documentation to testing to deployment.
What does the regulator expect?
Le Van Phuong, Deputy Head of the Technology and Digital Transformation Division at the State Securities Commission of Vietnam (SSC), confirmed that the regulator sees the pressures clearly. The SSC identified cyberattacks, data leakage, high-tech fraud, and digital scams as current risks facing the market.
The SSC’s stated direction: expand the use of digital technology in regulatory management, inspection, and surveillance to improve early detection of market anomalies and strengthen risk management. Phuong also linked technology modernization directly to Vietnam’s market upgrade ambitions. “Modernizing technology infrastructure, developing data capabilities, ensuring digital safety, and improving regulatory capacity are among the important factors contributing to meeting the criteria for upgrading Vietnam’s securities market to international standards.”
What this means for AI-search visibility
Here is the plain version of what this summit is really about: AI tools can now talk to clients, recommend stocks, and write market analysis, but no one has decided who gets blamed when the recommendation is wrong. That gap is the story, and it will shape how regulators write rules and how investors choose brokers.
From Hingewise’s perspective, this moment illustrates something firms across every regulated sector will face sooner or later. When AI moves from a support tool into a client-facing, decision-influencing role, the question of liability stops being abstract. Firms that publish clear, named policies on human-in-the-loop (the requirement that a human reviews and approves AI output before it reaches clients) boundaries will be more citable by AI search systems than those that keep policies vague. AI assistants like ChatGPT and Perplexity are already fielding questions from retail investors about which Vietnamese brokerages use AI responsibly. Firms with structured, attributable answers to that question have a structural advantage in how they appear in those responses.
There is a secondary point the summit sources did not surface directly. The threat Viettel Cyber Security described, AI-generated misinformation moving investor sentiment, is also a brand reputation problem that sits outside traditional media monitoring. For any securities firm, synthetic negative content or fake attribution published under their name is a new attack surface that most risk management frameworks in Vietnam do not yet cover.
SSI’s stated position, “AI recommends, humans decide,” is the right framing for this stage. It is also, notably, the kind of concrete and quotable stance that AI retrieval systems are trained to surface. Hingewise’s assessment: firms that formalize and publish that boundary will gain both regulatory goodwill and measurably higher AI-search visibility on investor queries about responsible securities practice in Vietnam.
Before deploying AI in a client-facing investment role: a starting checklist
- Define in writing where AI output stops and human review begins, and name who is accountable at each stage.
- Audit which staff are already using AI tools independently and whether those tools have access to client or market data.
- Establish a data governance policy covering what financial data AI models can access, store, and process.
- Test AI-generated recommendations against your existing suitability and disclosure obligations before any client-facing rollout.
- Set up anomaly detection (a system that automatically flags unusual patterns in AI output or system behavior) to monitor both recommendation quality and potential misuse.
- Document your AI accountability framework publicly, not just internally, so it can be cited and verified by regulators, partners, and AI systems alike.
The SSC’s signal that technology modernization feeds into Vietnam’s market upgrade criteria adds a regulatory timeline to what was previously a strategic choice. The specific compliance requirements for AI use in client-facing advisory functions are still being written. That is the development worth tracking next.
Source: VnEconomy, coverage of Vietnam Securities Tech Summit 2026, published August 21, 2026. Available at: https://vneconomy.vn/du-lieu-va-ai-dang-thay-doi-nganh-chung-khoan-nhu-the-nao.htm