Claude Code vulnerability exploited by hackers through phishing, encrypting users and making them targets of attack

Security risks strike again. Recently, the security research community exposed a serious vulnerability in Anthropic’s Claude Code tool, which allows hackers to execute arbitrary commands completely bypassing user authorization, with actual attack cases already reported.

High-risk privilege escalation vulnerability being exploited by hackers

This CVE-2025-64755 vulnerability involves two critical security issues: privilege escalation and code execution. Most seriously, attackers can execute commands directly on the system without any user confirmation. Researcher 23pds from SlowMist shared a detailed report from international security researcher Adam Chester, confirming the authenticity of this vulnerability.

What makes this vulnerability especially dangerous is that the full proof-of-concept (PoC) code has been publicly released online, meaning any hacker with basic technical skills can easily access attack tools. According to tracking, professional hacking groups have already begun exploiting this vulnerability to target cryptocurrency users with phishing attacks.

Hackers are targeting crypto users with targeted attacks

The reason the crypto space has become a priority target for hackers is that, once successful, victims’ digital assets can be transferred instantly. Researcher 23pds specifically pointed out that there is evidence showing hackers are using this vulnerability to conduct phishing activities against crypto users. Attackers often impersonate legitimate applications to trick users into running code snippets containing malicious commands.

Similar threats have also appeared in Cursor tools

It is worth noting that this is not the first time such serious vulnerabilities have appeared in code editing tool ecosystems. Previously, the popular Cursor tool also exposed similar privilege escalation and command execution vulnerabilities. Both exploit similar principles, mainly leveraging flaws in permission management to execute unauthorized commands. This highlights the need for greater security focus in AI-assisted programming tools.