KARACHI: Binance says its artificial intelligence systems stopped about $4.6 billion in potential losses in the first half of 2026 and shielded more than eight million users, a claim that lands as Pakistan builds rules for a crypto market it has only recently begun to regulate.
The exchange said its systems intercepted millions of scam and phishing attempts between January and June, blacklisted more than 42,000 malicious addresses and sent out over 14,000 real-time risk warnings a day.
More than 100 AI models now sit behind its anti-fraud controls. They make 80 to 90% of real-time risk decisions, with human analysts handling complex and high-risk cases, setting thresholds and retraining the models as new scam patterns appear.
The technology runs through identity verification, account security, payments and transaction monitoring. Binance said AI-enabled know-your-customer checks have made some workflows up to 100 times more efficient, and that more than 24 AI projects are live across compliance and onboarding.
Its wallet arm reported separate numbers. The Binance Wallet Security Center helped users avoid roughly $540 million in losses, filtered about 206 million spam transfers and flagged close to 4.93 million high-risk transactions across 19 networks.
Chief Security Officer Jimmy Su said the company keeps pairing the scale of AI with human expertise against fraud and social engineering. Defences cannot simply be assumed to work and have to be challenged, he said of the firm’s attack simulations.
The headline figure needs a caveat. It is Binance’s own estimate of losses avoided, not money recovered, and the company gave no breakdown by country, product or type of scam. As a last line of defence it points to its Secure Asset Fund for Users, set up in 2018 and holding about $1 billion.
For Pakistan the numbers are more than corporate housekeeping. The country built a virtual assets regime in six months, by its crypto chief’s account, and online deception is already a courtroom staple, from a fake Barclays advertisement that cost victims Rs19.39 million to a counterfeit NADRA website busted in Lahore.
Phishing, impersonation and fake investment offers are the same tricks Binance says its models are trained to catch. How well they do so for Pakistani users is something the company’s figures do not yet show.
Leave a Reply