Description

Bitdefender researchers uncovered a dangerous cyberattack aimed at developers and AI users searching for Claude Code, a tool linked to Anthropic’s Claude AI model. Attackers placed fake Google sponsored ads at the top of search results, making them appear trustworthy and official. When users clicked these ads, they were redirected not to Anthropic’s real website, but to a fake documentation page hosted on a Squarespace subdomain. The page looked almost identical to the real Claude documentation, copying its design, layout, and navigation, which made the scam very hard to spot. The attack relies on a deceptive social?engineering method called ClickFix, which tricks users into running harmful commands themselves. Instead of automatically infecting systems, the fake site convinces users to manually copy and run terminal commands, believing they are installing legitimate software. The commands are actually malicious. This approach works well because developers often trust terminal instructions from documentation pages and do not expect harm from seemingly official sources. The malware behaves differently depending on the user’s operating system. On Windows, victims are told to run a command using a built?in Microsoft tool, which downloads malware that steals saved passwords and credentials. On macOS, users run a hidden, complex command that installs a backdoor program without any visible warning. Once active, attackers can remotely control the infected system, access files, and run commands silently. To avoid detection by Google, the attackers likely used a previously trusted advertising account belonging to a real company. Google has since shut down that account. This incident highlights the need for developers to be cautious, even when clicking sponsored search results or following online documentation.