Friday, November 6, 2026

Hackers Use Google Ads and Claude AI Chats to Steal macOS Credentials and Crypto Wallets

Threat actors have exploited Google Ads and Anthropic’s Claude shared-chat feature to distribute the MacSync Stealer to macOS users.

They used a social engineering technique called ClickFix, designed to steal credentials, browser data, cloud keys, sensitive files, and cryptocurrency wallets.

According to Zscaler Threat Hunting, this campaign took place between June 12 and June 19, 2026, and the malicious Claude chats became inaccessible after the company alerted Anthropic.

Google Ads and Claude AI Chats to Steal macOS Credentials

The campaign began when a user searched Google for terms like “Claude download,” “Claude AI,” “Claude code,” or “Claude Mac.” Attackers purchased advertisements that redirected victims to a legitimate Claude shared chat URL, creating a strong illusion of trust because the content was hosted on the official Claude domain.

The chats were reportedly branded as being “Shared by Apple Support,” likely through an attacker-controlled display name. They instructed visitors to paste a Terminal command to install or repair Claude.

Rather than delivering a conventional malicious application, the operation relies on ClickFix. This technique manipulates victims into manually pasting and executing commands.

MacSync Stealer ClickFix instructions hosted in a shared Claude chat (Source: Zscaler)
MacSync Stealer ClickFix instructions hosted in a shared Claude chat (Source: Zscaler)

The command uses curl and Base64 obfuscation to retrieve a malicious shell script from the attackers’ infrastructure, then pipes its output directly into zsh.

The initial script Base64-decodes and gzip-decompresses a second-stage payload before executing it using eval, providing attackers with an efficient multi-stage infection path while obscuring the final destination from casual inspection.

The second-stage script suppresses visible output by redirecting it to /dev/null, then fetches the core payload through a `/dynamic?txd=` endpoint using a hard-coded API key.

The downloaded content is piped to osascript, a method that minimizes the payload’s on-disk footprint. After completing the theft, the malware packages the collected data into `/tmp/osalogging.zip` and uploads it in 10 MB chunks via HTTP PUT. It can retry failed uploads up to eight times before deleting the archive to limit forensic traces.

MacSync Stealer seeks extensive access to user data. It tries to access the Cookies directory and, if blocked, modifies the user’s `.zshrc` file to establish persistence whenever Terminal is opened.

It also displays a deceptive “Full Disk Access required!” prompt. It opens the macOS Security & Privacy settings and attempts to persuade victims to grant the necessary permissions for broader data collection. If access is granted, the malware removes its persistence command and proceeds to steal data.

The information stealer copies macOS Keychain databases, Chromium and Gecko browser credential artifacts, password-manager extension data, shell histories, SSH, AWS, and Kubernetes configuration material, Telegram desktop files, as well as documents from the Desktop, Downloads, and Documents folders.

Example Zsh script returned by the example first-stage URL distributing MacSync Stealer. (Source: Zscaler)
Example Zsh script returned by the example first-stage URL distributing MacSync Stealer. (Source: Zscaler)

It specifically searches for extensions related to web-based cryptocurrency wallets. It collects data from desktop wallet applications into separate folders.

The malware also employs a fake macOS password prompt to capture credentials. It gathers information about the system, hardware, graphics, processes, and applications.

Zscaler identified Russian-language comments in the third-stage AppleScript, suggesting that the operator may be Russian-speaking.

Researchers also observed the adversary using a large cluster of domains styled as U.S. local-service websites, likely to make the staging infrastructure appear less suspicious.

Organizations are advised to block known domains and monitor for unusual activity such as curl commands in zsh, modifications to .zshrc, temporary archive creation, and outbound HTTP PUT activity from macOS endpoints.

Indicators of Compromise

TypeIndicator
First-stage deliverylasvegaslaminateflooring[.]com/curl/0e17984a73d0b1c9c7c3916d32c49c8937f2e42d4c72c543c82999463a507abb
Core payload endpointlasvegaslaminateflooring[.]com/dynamic?txd=0e17984a73d0b1c9c7c3916d32c49c8937f2e42d4c72c543c82999463a507abb
Wallet-targeted payloadlasvegaslaminateflooring[.]com/ledger/0e17984a73d0b1c9c7c3916d32c49c8937f2e42d4c72c543c82999463a507abb
Potential Ledger Live payloadlasvegaslaminateflooring[.]com/ledger/live/0e17984a73d0b1c9c7c3916d32c49c8937f2e42d4c72c543c82999463a507abb
Potential Trezor payloadlasvegaslaminateflooring[.]com/trezor/0e17984a73d0b1c9c7c3916d32c49c8937f2e42d4c72c543c82999463a507abb
Temporary collection archive/tmp/osalogging.zip
Persistence artifactModification of ~/.zshrc
Detection namesHTML.Trojan.ClickFix; OSX.PWS.MacSync

Note: IP addresses and domains are intentionally defanged (e.g., [.]) to prevent accidental resolution or hyperlinking. Re-fang only within controlled threat intelligence platforms such as MISP, VirusTotal, or your SIEM.

 Strengthen Your SOC by Accelerating Threat Detection & Rapid Investigations. -> Integrate ANY.RUN With Your SOC Now.

Divya
Divya
Divya is a Senior Journalist at GBhackers covering Cyber Attacks, Threats, Breaches, Vulnerabilities and other happenings in the cyber world.

Hot this week

How To Access Dark Web Anonymously and know its Secretive and Mysterious Activities

What is Deep Web The deep web, invisible web, or...

How to Build and Run a Security Operations Center (SOC Guide) – 2023

Today’s Cyber security operations center (CSOC) should have everything...

Russian Hackers Bypass EDR to Deliver a Weaponized TeamViewer Component

TeamViewer's popularity and remote access capabilities make it an...

Web Server Penetration Testing Checklist – 2026

Web server pentesting is performed under three significant categories: identity,...

ATM Penetration Testing – Advanced Testing Methods to Find The Vulnerabilities

ATM Penetration testing, Hackers have found different approaches to...

Storm-2603 Hackers Exploit SharePoint and Abuse Cloudflare Tunnel to Hide Remote Access

A December intrusion to Storm-2603, the China-based group Microsoft...

New GhostMap Android Vulnerability Can Crash Phones Permanently Without Opening an App or Granting Permissions

Bitdefender researchers have revealed a vulnerability called GhostMap that...

ARTEX AI Framework Linked to Bank Breaches and Cloud Attacks Found Exposed Across Hundreds of Servers

An autonomous penetration-testing framework linked to South Korean banking...

OpenMatter Network Positions Decentralized Computing as an Alternative to Data Centers for AI

Melbourne, Florida, November 5th, 2026, CyberNewswire OpenMatter Network, a pioneer...

Related Articles

Recent News