One and a half years after the theft of $1.46 billion attributed to North Korea’s Lazarus Group, Bybit has released figures on its AI-powered security system. From January 1 to June 15, it blocked more than 30,000 suspicious withdrawal requests, which, according to the exchange, spared around 20,000 users from potential losses totaling over $700 million. The initial screening of a flagged transaction took an average of 4.7 minutes. Additionally, the AI linked approximately $212 million to fraudulent activities and blacklisted over 10,000 addresses.
Here, it’s important to note: “potential losses” refer to halted withdrawals, not confirmed thefts, and these figures cannot be independently verified. A significant portion of the blocked transactions were almost certainly legitimate.
Regarding code audits, the exchange claims that its AI identifies high-severity vulnerabilities roughly five times more frequently than manual analysis, and the time from system assessment to testing has been reduced from two weeks to just two hours. An automated red team scanned 1,489 publicly accessible infrastructure assets and discovered over a hundred critical vulnerabilities; in total, more than 100,000 alerts have been processed.