The risk infrastructure at the company covers abnormal trading activity, account takeover attempts, scams, and transaction fraud, supported by over 100 in-house AI models and continuous retraining by human risk analysts.

The privacy-first AI framework prioritizes data minimization, purpose limitation, and user rights safeguards, while balancing scalable protection with human oversight.

AI supports compliance and internal operations with more than two dozen initiatives across onboarding, screening, and partner due diligence, and an internal agentic tool has seen broad adoption across teams.

Social engineering defense uses multiple AI technologies, including computer vision to detect fake proof-of-payment images, with AI handling high-volume screening and humans providing validation.

Structured model governance guides the lifecycle of AI models—from development and validation to deployment and ongoing monitoring—to ensure sustained effectiveness.

KYC processes benefit from AI-enabled review pipelines that achieve significant efficiency gains in select workflows, while specialists handle higher-risk cases.

AI is embedded throughout the user journey—from identity verification to account security, payments, and transaction screening—enabling most checks to run automatically without slowing legitimate users.

A hybrid AI approach combines proprietary in-house models for platform-specific risks with external models for broader reasoning, plus an internal Red Team to test defenses.

AI-driven risk systems protected over 8 million users and prevented about $4.6 billion in potential losses in the first half of 2026.

In H1 2026, the company intercepted millions of scam and phishing attempts, blacklisted over 42,000 malicious addresses, and issued more than 14,000 real-time warnings daily, with 80–90% of real-time risk decisions powered by AI.

The company commits to ongoing investment in AI-driven scale and human judgment, including retraining models, adjusting thresholds, and involving experienced analysts for complex cases.