CoWorker, an AI Security Company, Addresses the "Invisible Risks" of AI Coding Agents. Launches "CoWorker AIDR" for Real-time Monitoring of Claude Code Execution, Offered Free of Charge.

CoWorker Inc. (Headquarters: Shinjuku-ku, Tokyo; Representative Director: Kazuki Yamasato; hereinafter referred to as "our company" or "CoWorker"), a developer and provider of AI security solutions, will offer the security scanner "CoWorker AIDR (AI Detection and Response)" free of charge to companies that inquire, starting from March 2026. This tool detects and defends against supply chain attacks when using the AI coding agent "Claude Code." This offer is for a limited time during the initial adoption phase.

This tool provides real-time monitoring of all tool calls made by AI, such as acquiring external libraries, browsing web pages, and generating/editing code. It automatically detects and blocks the import of contaminated packages, the injection of malicious code, and access to phishing sites.

Key Points of This Release (Summary)

・Addresses new cyber attack risks that are becoming apparent with the spread of AI coding agents.

・Monitors Claude Code's behavior to provide real-time defense against code tampering, information leakage, and phishing attempts.

・Aims to raise the security standard for AI development environments by distributing "CoWorker AIDR" to companies free of charge.

Background of the Free Release of "CoWorker AIDR"

AI Coding Agents as New Attack Vectors: Real-time Security Monitoring is Essential

AI coding agents dramatically improve developer productivity by autonomously performing tasks such as acquiring external libraries, referencing web pages, and generating/editing code. However, behind this autonomy lie supply chain risks such as the import of contaminated packages, the injection of malicious code, and redirection to malicious URLs.

Indeed, on March 24, 2026, the GitHub account of a main committer for the popular Python package "litellm" was compromised by the cybercriminal group "TeamPCP," resulting in a supply chain attack where malware was embedded in the package. This attack caused severe damage, with malicious code automatically executing every time the Python interpreter was launched, sending API keys, SSH keys, .env files, and credential information from local machines to external servers. Furthermore, our analysis revealed that this malware not only stole information but also possessed a backdoor function to download and execute arbitrary binaries from a C2 (Command and Control) server, and was equipped with a kill switch that allowed attackers to stop their activities and erase traces at any time. Organized and sophisticated tactics were employed, including posting hundreds of spam comments on GitHub Issues to obstruct discussions and reports.

Our company detected this incident early in our internal environment and prevented information leakage through rapid incident response. Based on this experience, we have come to recognize that real-time security monitoring is indispensable when using AI coding agents, leading to the free release of CoWorker AIDR...