CloudWalk Blocks Billions in Fraud with Proprietary AI System
With a powerful mix of machine learning and LLM-based agents, the fintech has built a dynamic fraud shield that detects suspicious behavior, disrupts fraud rings, and prevents fake accounts on its platform.
CloudWalk, the financial technology company behind InfinitePay and Jim.com, prevented over $3 billion (R$15 billion) in fraud losses in 2024, thanks to a proprietary AI-driven fraud prevention system. The multi-layered solution combines machine learning with recent advancements like LLM-based agents, each operating in a specific role. The system runs continuously, monitoring transactions in real time and adapting quickly to new fraud tactics—like fake accounts and money laundering schemes.
With over 99% accuracy, CloudWalk’s proprietary tech operates with no human intervention. It not only detects fraud attempts, but learns from each event, getting sharper with every interaction to better protect customer and partner funds.
“While fraud rings are using AI to create new attack models, we’ve built a robust, ever-evolving defense system powered by proven technologies,” says Alan Dias, Chief Risk and Compliance Officer at CloudWalk. “We’re talking about an extremely high trust index. Only 0.3% of the AI’s decisions require human review.”
A Network of Autonomous Alerts
A core component of CloudWalk’s risk and security infrastructure is a network of over 130 autonomous alerts. These tools combine machine learning with real-time behavioral rules based on both customer profiles and transaction history. When something abnormal is flagged, the system can ping the right team, trigger a temporary hold, or—in confirmed fraud cases—lock the account permanently. Coverage spans all major payment types, including card, Pix, boleto, and wire transfers, ensuring comprehensive risk mitigation.
The system has also drastically reduced transaction review times. Today, it takes just 0.05 seconds to analyze a payment—a task that could take a human up to 3 minutes. Since launch, this automation has saved an estimated 5,000 work hours, along with significant operational costs.
“Just looking at Pix transaction reviews alone, we’re saving about $1.2 million per month (R$6 million),” says Dias.
AI in the Fight Against Money Laundering
Beyond machine learning, CloudWalk has rolled out a set of internally developed AI agents over the past two years. One of the most strategic now focuses on anti-money laundering (AML). This agent scans hundreds of internal and external data sources—including Pix transfers, boleto payments, politically exposed person databases, arrest warrants, criminal records, and corporate affiliations.
When suspicious activity is detected, it can temporarily freeze accounts, request documents, and directly engage the customer.
“In this case, we saw an 80% reduction in analysis time because the agent handles the entire process end-to-end,” says Dias.
According to him, five agents are now running in the Risk area alone, delivering efficiency gains, cost cuts, and better security.
“These agents took over tasks previously handled by more than 100 analysts—who are now focused on building even more advanced fraud and AML defenses,” he adds.
