Advancing Equity and Accuracy in Deepfake Detection
Ant International has secured first place in the NeurIPS Competition of Fairness in AI Face Detection, a prestigious challenge held during the Conference on Neural Information Processing Systems — one of the most influential AI conferences globally. The competition attracted over 2,100 submissions from 162 teams across more than 40 countries, all vying to develop AI models that balance high performance with fairness across diverse demographic groups.
The core challenge addressed by the competition was the growing threat posed by deepfakes to facial recognition systems. As synthetic media becomes increasingly sophisticated, it undermines both the accuracy and equity of AI-driven identity verification. This is particularly critical in financial services, where even minor flaws in detection can lead to fraudulent transactions or discriminatory access denials.
A Novel Approach: Mixture of Experts Meets Adversarial Debiasing
Ant International’s winning model leverages a unique architectural design combining a Mixture of Experts (MoE) framework with an adversarial debiasing mechanism. This dual-network approach trains two competing sub-models simultaneously: one focused on detecting deepfake content, and another designed to actively suppress reliance on sensitive attributes such as skin tone, gender, or age.
This adversarial setup ensures the primary detection network learns to identify genuine indicators of manipulation — such as inconsistent lighting, unnatural motion, or texture anomalies — rather than relying on demographic patterns that may be underrepresented or misrepresented in training data.
Crucially, the model was trained on a globally representative dataset enriched with real-world payment fraud scenarios. This not only improves generalization but also strengthens its ability to perform equitably across different regions and populations.
Dr. Tianyi Zhang, General Manager of Risk Management and Cybersecurity at Ant International, emphasized the strategic importance of fairness: “A biased AI is an insecure AI. Our model’s commitment to fairness isn’t just an ethical imperative — it’s a security necessity. It prevents exploitation through deepfakes while ensuring reliable identity verification for every user, regardless of background.”
Real-World Applications in Financial Services
The technology is already being deployed within Ant International’s global ecosystem, enhancing security and compliance across its suite of financial products and services.
Account Protection Against Deepfake Threats
Ant International’s AI-powered anti-deepfake system now achieves a detection rate exceeding 99.8%, providing robust defense against account takeovers and digital impersonation. This level of precision is maintained consistently across 200 markets, ensuring equitable protection for users regardless of demographic profile.
By prioritizing fairness, the system avoids the pitfalls of algorithmic bias that could otherwise result in higher false rejection rates for certain groups — a common issue highlighted in NIST studies showing disproportionately elevated error rates for women and people of color in commercial facial recognition tools.
Strengthening eKYC Compliance
Electronic Know Your Customer (eKYC) processes are central to regulatory compliance in financial services. With the NeurIPS-winning technology, Ant International and its partner institutions can implement identity verification systems that meet stringent global standards without introducing algorithmic bias.
This is especially valuable in cross-border operations, where regulatory requirements vary widely. The fair and accurate detection system enables seamless customer onboarding while reducing the risk of discrimination or exclusion based on appearance.
Expanding Financial Inclusion
Equitable AI is foundational to Ant International’s mission of expanding access to financial services in emerging markets. By minimizing bias in identity verification, the company can serve broader, more diverse populations — including those historically underserved due to technological or infrastructural barriers.
This advancement supports inclusive growth, helping bridge gaps in financial access while maintaining high-security standards.
Building a Full-Cycle Security Framework
Ant International operates a vast global network serving over 150 million merchants and 1.8 billion user accounts through platforms such as Alipay+, Antom, Bettr, and WorldFirst. To protect this ecosystem, the company has embedded AI security into its core infrastructure.
AI SHIELD: A Foundational Risk Management System
Central to this effort is AI SHIELD, a comprehensive, full-lifecycle risk management framework built on the AI Security Docker platform. Designed to safeguard AI services from vulnerabilities such as unauthorized access, data leakage, and adversarial attacks, AI SHIELD reduces the risk of AI-related security incidents by up to 90% through advanced assessment protocols and continuous monitoring.
The framework provides end-to-end protection across the transaction lifecycle — from initial authentication to post-transaction monitoring. For example, the Alipay+ EasySafePay 360 solution, powered by AI SHIELD, has been shown to reduce account takeover incidents in digital wallet payments by 90%.
This integrated security posture underscores Ant International’s commitment to treating AI not just as a tool for innovation, but as a cornerstone of enterprise resilience.
The Broader Implications
The success at NeurIPS reflects a broader industry shift toward responsible AI development. As AI becomes more deeply embedded in financial systems, the demand for fairness, transparency, and robustness grows exponentially.
Ant International’s win demonstrates that technical excellence and ethical responsibility are not mutually exclusive — they are interdependent. High-performing AI must be fair to be secure, and secure AI must be fair to be trustworthy.
With its focus on global representation, adversarial fairness mechanisms, and real-world application, Ant International’s model sets a new benchmark for what is possible in AI-powered identity verification — one that balances cutting-edge performance with social responsibility.
About Ant International
With headquarters in Singapore and main operations across Asia, Europe, the Middle East and Latin America, Ant International is a leading global digital payment, digitisation and financial technology provider. Through collaboration across the private and public sectors, our unified techfin platform supports financial institutions and merchants of all sizes to achieve inclusive growth through a comprehensive range of cutting-edge digital payment and financial services solutions. To learn more, please visit https://www.ant-intl.com/
