Artificial IntelligenceCybersecurityNewswireTechnology

Beware Fake Claude Code Pages: “InstallFix” Attacks Surge

Originally published on: March 10, 2026
▼ Summary

– Attackers are tricking users with fake installation pages for Anthropic’s Claude Code AI tool, using cloned websites and malicious Google ads.
– The fake pages appear legitimate but contain instructions that download malware, specifically the Amatera Stealer for Windows and likely similar info-stealers for macOS.
– This “malvertising” attack is effective because it bypasses email security, as users initiate the search themselves for a tool they intend to install.
– Researchers have coined the term “InstallFix” for this tactic, which exploits the common practice of pasting installation commands from websites into terminals.
– The threat extends beyond Claude, as any popular, easily cloned tool or website is a potential target for this type of impersonation and malvertising.

Cybersecurity experts are raising the alarm about a sophisticated malware campaign targeting developers searching for the popular Claude Code AI tool. Researchers have identified a surge in fake installation pages that perfectly mimic the official Anthropic site, tricking users into downloading information-stealing malware instead of the legitimate software. These malicious pages are being promoted through paid Google ads, appearing at the very top of search results for terms like “install Claude Code” or “Claude Code CLI,” making them dangerously convincing.

The attackers operate by creating a near-identical copy of the authentic Claude installation guide and hosting it on a deceptive, lookalike domain. While most links on the fraudulent page redirect to the real Anthropic website, the critical installation instructions have been completely replaced. The provided command, which users are encouraged to paste directly into their terminal, is malicious. For Windows systems, it downloads a piece of malware known as Amatera Stealer; for macOS users, the command likely retrieves a similar info-stealing program designed for that operating system.

This method is particularly effective because it bypasses traditional email security filters. There is no phishing email to scrutinize or suspicious link in a message. The user initiates the entire interaction themselves by searching for a tool they genuinely want to install. Security analysts note that pasting commands from a website into a terminal has become a standard practice for developers, creating a perfect opportunity for this kind of attack. The researchers have dubbed this social engineering tactic “InstallFix,” drawing inspiration from the common “ClickFix” terminology.

The fake pages are so well-crafted that they are virtually indistinguishable from the real thing at a glance. As one researcher pointed out, unless a user meticulously examines the URL embedded within the one-line installation command, a practice few people follow, they would have no reason to suspect foul play. This campaign highlights a broader threat beyond a single AI tool; any popular software or website that attracts significant traffic and can be easily cloned is a potential target for this type of malvertising and impersonation.

Attackers are increasingly focusing on compromising online advertising accounts to hijack existing ad budgets and launch more malicious campaigns. This approach allows them to place deceptive ads directly in front of users who are actively seeking specific software, dramatically increasing the likelihood of a successful infection. The trend underscores the need for heightened vigilance, even when interacting with what appear to be official sources found through search engines. Always verify URLs carefully and consider downloading tools directly from the vendor’s verified website or official repository.

(Source: HelpNet Security)

Topics

malware distribution 95% social engineering 90% malicious advertising 88% phishing attacks 85% ai tool security 82% claude code 80% domain spoofing 78% search engine exploitation 75% info-stealing malware 73% developer tool risks 70%