Attackers Impersonate Anthropic's Claude Website to Spread PlugX Remote-Access Trojan

Claude Code is an Anthropic development tool that helps users write and modify code from a terminal. Because development environments often contain source code, API keys, cloud credentials and cryptocurrency wallet data, a successful attack using a fake installation page could also compromise corporate systems and supply-chain security.

The InstallFix attack disclosed on March 9 mimicked a Claude Code installation page and tricked users into copying and running a curl-to-bash command, circumventing the caution typically associated with downloads. The malware then installed Amatera Stealer to collect browser passwords, session tokens and development-environment credentials. No financial losses were publicly reported.

As generative AI services gain popularity, well-known brands such as Claude have become targets for impersonation in social-engineering attacks. Cybersecurity company Sophos said attackers cloned Anthropic’s Claude website and exploited users’ trust in AI tools to induce them to download software. No transaction value was involved, but victims’ computers could be remotely controlled.

Sophos recently found that hackers primarily used malicious advertising and SEO poisoning to direct search users to the fake Claude website, where they distributed a backdoor called Beagle. Once installed, Beagle can receive and execute attackers’ commands. As of July 20, 2026, available information had not disclosed when the campaign was first discovered, the number of victims or the amount of losses.