Users attempting to access Claude Code online are facing a significant security risk. Security firm Huntress has uncovered an attack campaign aimed at emptying the crypto wallets of macOS users through a fake server guide.

Attackers are using paid advertisements on Google to direct users to a fake server guide hosted on claude.ai. This guide creates a sense of trust because it operates under Anthropic’s own security certificate.

How does the MacSync malware work?

This six-step attack chain, named MacSync by Huntress, involves a remote access Trojan and an information thief. The attack becomes persistent on the system, starting with the user running a single command via Terminal.

In the most critical phase of the attack, Ledger and Trezor applications installed on the victim’s computer are targeted. The malware modifies these applications, redirecting the user to a fake recovery screen and hijacking wallet recovery terms.

Anthropic’s public chat sharing feature is exploited by attackers. Because the shared content is presented with Anthropic’s own certificate, no warning sign appears in the address bar.

Attackers mislead users by setting the chat screen name as “Apple Support”. Anthropic’s security banner also confirms this name, making the attack appear legitimate.

Warnings from security experts

It is known that AMOS malware was previously distributed through similar methods on platforms such as ChatGPT and Grok. This shows that shared guides for artificial intelligence tools are not always reliable.

Security experts emphasize the need to be cautious about online resources, even if they rank highly in Google ads. It is especially important to be wary of guides that ask you to run commands via the Terminal.

What are your thoughts on these complex attack methods?