Digital payments and fintech company Ant International has won the NeurIPS competition for fairness in AI facial detection. The company says it is focused on developing secure and inclusive financial services, especially as deepfake technology becomes more commonplace.
The increasing use of facial recognition in many fields has highlighted the issue of algorithmic bias in AI. A study conducted by NIST (National Institute of Standards and Technology) found that many widely used facial recognition algorithms have significantly higher error rates when analyzing the faces of women and people of color, and this disparity is due to a lack of diversity in the training data and demographics of the people building and controlling many mainstream AI platforms. The results of biased algorithms can lead to the denial of financial services to large sections of the population and are considered weaknesses in security protocols.
The NeurIPS competition was held in parallel with the acclaimed AI conference, the Conference on Neural Information Processing Systems, and challenged participants to create high-performance and fair AI models that cover a variety of demographic factors, including gender, age, and skin color. The Ant International team beat out more than 2,100 entries from 162 teams from around the world. The task given was to accurately detect 1.2 million AI-generated facial images that were selected to be reasonably representative of demographic groups.
The approach taken by Ant’s superior AI models combines a Mixture of Experts (MoE) architecture with bias detection mechanisms. This system trains two competing neural networks. One focuses on identifying deepfakes, while the other is designed to counter the former, forcing them to ignore demographic characteristics. This dynamic process ensures that the system learns how to detect true signs of manipulation, rather than carelessly relying on demographic patterns. The model was trained on globally representative datasets and incorporated real-world payment fraud scenarios to ensure performance at scale.
“Biased AI systems are inherently insecure,” explains Dr. Tianyi Zhang, general manager of risk management and cybersecurity at Ant International. “The fairness of our model is not just an ethical issue; it is fundamental to preventing deepfakes from being exploited and ensuring reliable identity verification for all users.”
The technology behind the winning entry is now being integrated into Ant’s payments and financial services to combat the threat of deepfakes, and the companies say it has achieved a detection rate of over 99.8% across all demographics and the 200 markets in which Ant operates.
Ant’s technology helps customers meet global electronic know-your-customer (eKYC) standards without algorithmic bias, especially during customer onboarding. This is considered to be particularly important in emerging markets where greater financial inclusion may be hindered.
Ant International serves more than 150 million merchants and 1.8 billion user accounts, known for services such as Alipay+, Antom, Bettr, and WorldFirst. The company says AI security is a pillar of its business. AI SHIELD is a risk management framework built on AI Security Docker that reduces the risk of AI service vulnerabilities such as unauthorized access and data leakage.
AI SHIELD powers a suite of risk management solutions that provide broad protection for financial transactions, including protection against deepfake attacks and fraud. The company says that Alipay+ EasySafePay 360 has reduced account takeover incidents in digital wallet payments by 90%.
(Image source: “Beautiful Portrait Abstract Art, Three-dimensional Shapes, Geometric Shapes, Neo-Tokyo Colors, Subdued Colors, Pixar, Art Station, Greg Rutkowski, Samdosart, GE” – Public Domain)
Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expo in Amsterdam, California, and London. This comprehensive event is part of TechEx and co-located with other major technology events. Click here for more information.
AI News is brought to you by TechForge Media. Learn about other upcoming enterprise technology events and webinars.
