Binance said its AI security systems blocked more than $10.5B in potential fraud losses over 15 months.
Binance said on May 11 that its AI-driven security infrastructure blocked more than $10.5 billion in potential user losses between January 2025 and the end of Q1 2026. The exchange deployed over 100 AI models across its systems to counter a surge in AI-powered attacks during that period. #Binance #CryptoSecurity #AI #CryptoFraud
Crypto-related fraud reached $17 billion globally in 2025, a 30% increase from the prior year, according to Binance's own research. The exchange said AI has lowered the cost and complexity of attacks. Smart contract exploits now cost attackers as little as $1.22 per contract, down 22% month-over-month.
AI Strengthens Attacks and Defense
Advanced AI models are achieving a 72.2% success rate in attack scenarios, Binance said. The exchange reported that 76% of AI-driven scams now rank at the highest tier for both scale and severity. Attackers are deploying deepfakes, voice cloning, phishing bots, and impersonation schemes across messaging platforms.
Binance said its AI systems contributed to a 60–70% reduction in card fraud rates compared to industry benchmarks. The company also launched Binance AI Pro, a product designed to limit risk at the architecture level. Under that framework, funds managed by AI agents are held separately from main user accounts and restricted to trading activity only, with no withdrawal access.
Approximately 12% of third-party tools submitted to Binance's marketplace have been flagged as potentially risky. The exchange reviewed more than 71,000 law enforcement requests in 2025. It helped authorities confiscate $131 million in illicit funds and recovered $12.8 million across 48,000 individual cases, a 41% year-over-year increase.
"AI is reshaping both sides of the security equation," Binance wrote. "It's making attacks more scalable, more convincing, and harder to detect, while also enabling a new generation of defenses that are faster, smarter, and more adaptive."
