Google Ads Campaign Delivers MacSync Through Claude Lures
In July 2026, a Google Ads malvertising campaign targeted macOS users searching for Claude Code installation help and delivered the MacSync infostealer through trusted-looking pages. Deriv reported that one live lure used a legitimate claude.ai/share page and a Base64-obfuscated shell command, while related reporting described Google Sites and shared-chat variants. Researchers reported credential, browser-session, developer-secret, and cryptocurrency-wallet theft capabilities.
A Google Ads malvertising campaign has targeted macOS users searching for Claude Code installation instructions, using trusted-looking pages and copy-paste Terminal commands to deliver the MacSync infostealer. Deriv's July 27 analysis reports that a sponsored result for "how to install Claude Code on Mac" led to a legitimate claude.ai/share page that presented an installation guide and concealed its download location in a Base64-obfuscated command.
The campaign relies on social engineering rather than a browser exploit or a compromise of Anthropic infrastructure, according to Deriv. The lure exploits a routine developer workflow: searching for a CLI installation command, trusting a sponsored result, and pasting the displayed command into Terminal.
Multiple Claude-themed delivery variants
Reporting on MacSync has documented more than one delivery configuration. Deriv identified a Claude shared-page lure hosted on claude.ai, while Beelzebub Labs research summarized by CybersecurityNews and CyberPress described a related campaign that used sponsored ads leading to a fake Claude Code page on Google Sites.
GBHackers, citing Zscaler Threat Hunting, reported a June 12-19 campaign in which attacker-controlled Claude shared chats were branded as "Shared by Apple Support." According to that report, the malicious chats became inaccessible after Anthropic was alerted. The shared-chat content instructed users to paste a command that retrieved a payload with curl, decoded Base64 data, and piped output to zsh.
These reports describe a common delivery pattern, but they should not be treated as evidence that every Claude-themed ad uses the same infrastructure or payload build. The current Deriv report specifically identifies a claude.ai/share lure, whereas earlier analyses identify Google Sites pages and other shared-chat content.
What the payload targets
Deriv characterizes MacSync as a macOS infostealer designed to collect credentials, browser sessions, developer secrets, and cryptocurrency wallet data. Gridinsoft's summary of Beelzebub Labs research lists browser profiles, macOS Keychains, SSH keys, AWS credentials, Kubernetes configuration files, shell history, Telegram sessions, Apple Notes, and wallet extensions among the reported targets.
GBHackers' account of the Zscaler research adds technical detail on the staged execution chain. It reports that the second-stage script suppresses output, retrieves a core payload through a /dynamic?txd= endpoint, executes downloaded content through osascript, packages collected data into /tmp/osalogging.zip, and uploads the archive in chunks. Those implementation details are reported behavior, not an indication that all MacSync samples or campaigns operate identically.
Earlier Beelzebub Labs reporting, as summarized by CyberSecurityNews and CyberPress, also described a fake password prompt and a second stage that modified Ledger application files to collect cryptocurrency recovery phrases. That capability is particularly consequential for affected users, but it is tied to the campaign samples analyzed by those researchers.
A developer-tool trust problem
The immediate risk is not a vulnerability in Claude Code itself. It is the combination of paid-search placement, legitimate or widely trusted hosting domains, plausible installation documentation, and user-executed shell commands. Deriv reports that the latest lure used legitimate infrastructure specifically to make conventional domain checks less useful.
For security teams, this campaign illustrates a broader pattern in which software-install instructions become the attack surface. Organizations that support developer workstations commonly treat pasted shell commands, cloud credentials, SSH keys, browser sessions, and wallet extensions as distinct detection and response concerns. The reported MacSync collection set combines several of those assets in one endpoint compromise, increasing the value of monitoring unexpected shell execution and outbound archive uploads from developer Macs.
Users seeking Claude Code should obtain installation commands from Anthropic's official documentation rather than sponsored search results or shared pages. Anyone who executed a suspicious command should treat stored browser sessions, developer credentials, SSH keys, cloud keys, and cryptocurrency wallet data as potentially exposed, based on the capabilities reported by the cited researchers.
Key Points
- 1Google Ads and trusted Claude-hosted pages can make a malicious Terminal instruction appear credible to developers seeking legitimate CLI setup guidance.
- 2Reported MacSync collection targets include browser sessions, Keychain data, SSH keys, cloud credentials, Kubernetes configuration, and cryptocurrency wallet information.
- 3Comparable social-engineering campaigns shift defensive attention from malicious downloads alone to copied shell commands and credential exposure on developer endpoints.
Scoring Rationale
The campaign targets AI-tool users through a credible installation workflow and reportedly seeks high-value developer credentials, browser sessions, and cloud configuration data. It is a notable endpoint-security threat for software teams, although the reported activity is a campaign rather than a vulnerability in Claude Code or Anthropic infrastructure.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems

