Claude Used to Create Malicious npm Packages: Over 670 Compromised Threaten Open Source
A recent cybersecurity incident reveals how large language models (LLMs) are being weaponized for malicious software development. Security researcher Sibi Moosa spotted an attacker using the alias "mousie-5212-super-formatter" leveraging Anthropic's Claude AI to generate harmful code and contaminate the npm package ecosystem. Within a short timeframe, over 670 malicious packages were uploaded to the npm registry, raising alarms over the speed and automation of such attacks.

At the core of this attack is the use of AI to drastically reduce the effort required to create malicious code. The compromised npm packages target developer credentials like npm tokens and GitHub tokens, as well as source code from private GitHub repositories. The attacker employs Claude to craft coherent data-stealing scripts and exfiltrates the stolen information to their own repositories. This case underscores how generative AI, while boosting productivity, also serves as a force multiplier for attackers, amplifying both efficiency and automation.
Experts note that automated package poisoning and code theft through AI models signify a new, intelligent stage in supply chain attacks. Traditional signature-based defenses are ill-equipped to counter the highly variable and deceptive malicious payloads that AI can generate. With the growing adoption of AI coding assistants, preventing their misuse for vulnerability exploitation and malware creation has become a critical priority in AI security governance.
Related article
Ando Launches Team Messaging App Bridging Human and AI Collaboration
In 2025, as Sara Du assisted companies in developing MCP servers, a recurring question emerged: how could teams leverage AI agents directly within Slack? The path was obstructed by significant technical barriers, including the complexity of message s
Sequoia-backed Empirik raises $21M to forecast outages before they strike
Before joining Sequoia Capital in 2020 as chief digital and information officer, Avon Puri spent over a decade running infrastructure at Rubrik and VMware. Three years ago, as large language models began showing their true potential, Puri, alongside
OpenAI launches GPT-5.5 Instant as ChatGPT’s new default model
OpenAI launched GPT-5.5 Instant on Monday, replacing GPT-5.3 Instant as the default ChatGPT model. The company states that this update minimizes hallucinations in high-stakes fields like law, medicine, and finance, while preserving the fast response
Related Special Topic Recommendations
Comments (0)
0/500
A recent cybersecurity incident reveals how large language models (LLMs) are being weaponized for malicious software development. Security researcher Sibi Moosa spotted an attacker using the alias "mousie-5212-super-formatter" leveraging Anthropic's Claude AI to generate harmful code and contaminate the npm package ecosystem. Within a short timeframe, over 670 malicious packages were uploaded to the npm registry, raising alarms over the speed and automation of such attacks.

At the core of this attack is the use of AI to drastically reduce the effort required to create malicious code. The compromised npm packages target developer credentials like npm tokens and GitHub tokens, as well as source code from private GitHub repositories. The attacker employs Claude to craft coherent data-stealing scripts and exfiltrates the stolen information to their own repositories. This case underscores how generative AI, while boosting productivity, also serves as a force multiplier for attackers, amplifying both efficiency and automation.
Experts note that automated package poisoning and code theft through AI models signify a new, intelligent stage in supply chain attacks. Traditional signature-based defenses are ill-equipped to counter the highly variable and deceptive malicious payloads that AI can generate. With the growing adoption of AI coding assistants, preventing their misuse for vulnerability exploitation and malware creation has become a critical priority in AI security governance.
Ando Launches Team Messaging App Bridging Human and AI Collaboration
In 2025, as Sara Du assisted companies in developing MCP servers, a recurring question emerged: how could teams leverage AI agents directly within Slack? The path was obstructed by significant technical barriers, including the complexity of message s
Sequoia-backed Empirik raises $21M to forecast outages before they strike
Before joining Sequoia Capital in 2020 as chief digital and information officer, Avon Puri spent over a decade running infrastructure at Rubrik and VMware. Three years ago, as large language models began showing their true potential, Puri, alongside
OpenAI launches GPT-5.5 Instant as ChatGPT’s new default model
OpenAI launched GPT-5.5 Instant on Monday, replacing GPT-5.3 Instant as the default ChatGPT model. The company states that this update minimizes hallucinations in high-stakes fields like law, medicine, and finance, while preserving the fast response





Home






