Page 66 - ISC PROCEEDINGS 21.4

A distinctive feature of the French model is the strong linkage between
technological innovation and the legal framework for data protection, particularly in
compliance with the EU’s General Data Protection Regulation (GDPR). AI systems in the
public sector must ensure explainability, non-discrimination, and independent oversight
mechanisms. This approach reduces the risk of “algorithmic black boxes” and strengthens
accountability.
3.3.4. Comparison and policy implications for Vietnam
A comparison of the three models reveals several structural commonalities: public
financial data is almost fully digitized; management systems are integrated rather than
fragmented across agencies; risk-based analytics are applied to prioritize oversight; and
technology is closely linked to transparency and accountability.
However, the focus of reform differs across countries. Estonia prioritizes
interoperable data architecture; South Korea concentrates on building a large-scale
integrated ST system; while France emphasizes risk governance and AI ethics.
For Vietnam, these experiences suggest several important directions: developing an
interoperable public financial data architecture linking ST, public investment, and
procurement systems; establishing an integrated state budget management system
capable of near real-time monitoring of PIE; applying risk-based oversight with AI as a
decision-support tool; and strengthening the legal framework on data protection and
algorithmic transparency.
International experience demonstrates that BDA and AI are not standalone
solutions, but rather value-added layers built upon institutional reform and robust data
governance foundations. This constitutes the essential condition for Vietnam to move
toward an intelligent, transparent, and sustainable public financial governance model in
the digital era.
3.4. Challenges in implementing a public investment expenditure monitoring
model based on big data and AI
Although BDA and AI create significant opportunities to enhance efficiency,
transparency, and proactiveness in monitoring PIE, their practical implementation faces
several structural challenges. Three prominent groups of issues include: (i) data security
and system safety; (ii) ethical risks and algorithmic bias; and (iii) accountability in a data-
and AI-driven decision-making environment.
3.4.1. Data security and system security
A Big Data-based public investment monitoring system requires the integration of
large volumes of sensitive data, including ST information, contracts, payment records,
enterprise data, procurement data, and even geospatial and on-site imagery data.
According to IBM’s 2023 report, the global average cost of a data breach reached
approximately USD 4.45 million per incident - the highest level ever recorded. Notably,
the public sector and critical infrastructure rank among the sectors with the highest levels
of damage due to the sensitivity of their data.
When ST, public investment, and procurement systems are interconnected, the
risks of cyberattacks and unauthorized access increase exponentially. According to
statistics from the World Economic Forum, more than 95% of cybersecurity incidents
involve elements of human error or weak access governance. This indicates that the
challenge lies not only in technical solutions (encryption, multi-factor authentication,
access control), but also in the design of data governance mechanisms and internal
control systems.

65