Bybit Exposes macOS Malware Campaign Targeting Claude Code Searchers

Bybit's Security Operations Center (SOC) has lifted the lid on a multi-stage macOS malware campaign aimed squarely at developers googling "Claude Code"—Anthropic's AI-powered coding assistant. Think of it as the blockchain industry's contribution to the cybersecurity horror genre: AI meets malware meets SEO tricks. The campaign, spotted in March 2026, weaponized search engine optimization (SEO) poisoning to push a malicious domain above legitimate results, complete with a phishing page that could pass for the real thing at a glance.

The attack pulled off a classic two-stage malware chain. First up: a Mach-O dropper that planted an osascript-based infostealer—cousin to the notorious AMOS and Banshee families—running a multi-phase obfuscation routine to slurp browser credentials, macOS Keychain entries, Telegram sessions, VPN profiles, and crypto wallet data. Bybit researchers flagged targeted hits on more than 250 browser-based wallet extensions alongside several desktop wallet apps. The encore? A second-stage C++ backdoor sporting sandbox detection, encrypted runtime config, and persistence via system-level agents, complete with remote command execution through HTTP polling. Because why settle for stealing one thing when you can establish a full digital squatting operation?

The social engineering menu included fake macOS password prompts designed to validate and cache credentials—the digital equivalent of handing your keys to someone who says they're from "tech support." In some cases, attackers swapped out bona fide wallet apps like Ledger Live and Trezor Suite with trojanized versions sitting on malicious servers. The malware cast its net wide: Chromium browsers, Firefox variants, Safari data, Apple Notes, and those "definitely-not-a-password-document" files people keep naming things like "important_bank_stuff.txt."

Bybit's SOC flexed AI-assisted workflows across the entire malware analysis pipeline, cutting deep inspection of the second-stage backdoor from an estimated six to eight hours down to under 40 minutes. AI-powered reverse engineering and control-flow analysis meant detection measures shipped same day, while AI-generated reporting drafts cut turnaround time by roughly 70% compared to traditional methods. "Using AI to defend against AI is an inevitable trend," noted David Zong, Head of Group Risk Control and Security at Bybit. "We will further increase our investment in AI for security, achieving minute-level threat detection and automated, intelligent emergency response." Apparently, the robots are now eating other robots for breakfast.

The malicious infrastructure got flagged on March 12, with full analysis, mitigation, and internal detection wrapped up same-day. Public disclosure followed on March 20, along with detailed detection and remediation guidance so others could spot and neutralize similar threats. All associated domains and command-and-control endpoints have since been defanged for public consumption—because nobody likes a malware campaign that still has working URLs.