PulseAugur
EN
LIVE 07:28:42

GhostApproval vulnerability impacts 6 AI coding assistants, bypassing security checks

A newly disclosed vulnerability, dubbed GhostApproval, affects at least six AI coding assistants, potentially allowing malicious code to bypass security checks and gain system access. The flaw exploits symlink vulnerabilities to trick users into approving actions that could write to sensitive files like SSH keys. While some assistants from AWS, Cursor, and Google have been patched, others from Anthropic, Augment, and Windsurf remain vulnerable. AI

IMPACT This vulnerability could allow attackers to gain unauthorized system access through AI coding tools, highlighting the need for enhanced security in AI-powered development environments.

RANK_REASON Disclosure of a security vulnerability in AI coding assistant products.

Read on Mastodon — fosstodon.org →

AI-generated summary · Google Gemini · from 3 sources. How we write summaries →

GhostApproval vulnerability impacts 6 AI coding assistants, bypassing security checks

COVERAGE [3]

  1. Mastodon — fosstodon.org TIER_1 Español(ES) · [email protected] ·

    Wiz discovered GhostApproval: an attack that deceives AI assistants like Claude Code, Cursor, and Amazon Q Developer using symlinks (yes, the Unix vuln from the 70s)

    Wiz descubrió GhostApproval: un ataque que engaña a asistentes de IA como Claude Code, Cursor y Amazon Q Developer usando symlinks (sí, la vuln de Unix de los 70s).El agente ve un archivo local inofensivo, pero el symlink apunta a /etc/passwd o ~/.ssh/authorized_keys. El diálogo …

  2. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    📣🚨 # GhostApproval symlink vulnerabilities in major AI coding assistants could hide sensitive file targets, bypass approval checks, and enable system access too

    📣🚨 # GhostApproval symlink vulnerabilities in major AI coding assistants could hide sensitive file targets, bypass approval checks, and enable system access too. Read: https:// hackread.com/ghostapproval-fla ws-ai-coding-tools-outside-workspace/ # CyberSecurity # AI # Vulnerabili…

  3. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    Wiz disclosed GhostApproval: a symlink flaw in 6 AI coding assistants. A malicious repo can hijack the human-in-the-loop confirmation dialog to write to ~/.ssh/

    Wiz disclosed GhostApproval: a symlink flaw in 6 AI coding assistants. A malicious repo can hijack the human-in-the-loop confirmation dialog to write to ~/.ssh/authorized_keys. AWS, Cursor, and Google patched; Anthropic, Augment, Windsurf have not. https:// go.aintelligencehub.co…