Skip to content

AI security

Miasma attack targets AI assistant configurations

Written by an AI model from the linked sources and checked against them by a second model. How we work

All about: Claude Code versions and changes →
Image: scienspire.com

In June 2026, GitHub disabled 73 Microsoft repositories due to the Miasma malware. The virus targets Claude Code, Gemini CLI, Cursor, and VS Code. Unlike traditional attacks, the code runs when a project folder is opened, not when packages are installed.

Attackers use config files that AI assistants read automatically. In Claude Code and Gemini CLI, the malicious code is hidden in SessionStart hooks. In Cursor, the attack is done via prompt injection in a rules file, and in VS Code, via a modified tasks.json.

The program steals AWS, GCP, and Azure keys, GitHub Actions secrets, and data from password managers. Developers are advised to check for suspicious folders in repositories. If compromise is suspected, revoke all SSH keys, GitHub tokens, and cloud credentials.

Anthropic

Anthropic tests AI exploit development speed

Anthropic’s Mythos Preview model can turn a security patch into a working exploit in hours. Tests on Firefox and Windows show the model autonomously finds vulnerabilities faster than companies can distribute updates.

No comments yet. Be the first.

Join the conversation

Sign in to leave a comment. It appears after moderation.