Radware researchers scanned over 100,000 AI agent instruction files across 45,000 public GitHub repositories, finding hundreds containing real credentials and 60% of testable keys still live.
Public source
Publisher name
Public post
AI coding agents have created a new place for credentials to leak. Radware researchers scanned ~100,000 AI agent instruction files across ~45,000 public GitHub repositor…
Company
Radware
- Industry
- Computer and Network Security
- Location
- Tel Aviv, IL
- Company size
- 1,001–5,000 employees
About Radware
Radware® is a global leader in application security and delivery solutions for multi-cloud environments. The company’s cloud application, infrastructure, API, and AI security solutions use AI-driven algorithms for precise, behavior-based, real-time protection against sophisticated web, application, and DDoS attacks, API abuse, business logic threats, and malicious bots. Radware delivers end-to-end API security, including discovery, posture management, testing, and runtime protection, along with advanced protection for AI agents and models. Enterprises and carriers worldwide rely on Radware to address evolving cyberthreats, protect their brands and business operations, and reduce costs. For more information, please visit the Radware website.
See moreLatest activity
Latest activity from Radware
39 signals
Presence & Recognition
Radware and Dataiku are hosting a 45-minute webinar on how enterprises can move agentic AI from experimentation to production while maintaining visibility, governance, compliance, and runtime security on September 8, 2026
Presence & Recognition
Radware was named a Leader in the 2026 QKS Group SPARK Matrix for AI-Powered Application Security
Presence & Recognition
Radware is hosting a 45-minute webinar on how enterprises can move agentic AI from experimentation to production while maintaining visibility, governance, compliance, and runtime security on September 8, 2026, featuring Dror Zelber and Stephen Wagner.
Discover more
Similar signals
Similar public activity from other companies.
Research & Knowledge
Cognni
Cognni published a research study analyzing over 17,000 third-party AI agent skills on the largest open-source marketplace, finding 520 leaking credentials with almost 90 percent immediately exploitable.
Research & Knowledge
Veracode
Veracode research across more than 100 AI models found an average security pass rate of roughly 56%, meaning about 44% of code-generation tasks introduced a known vulnerability.
Research & Knowledge
Wiz
Wiz analyzed 90 days of attack telemetry across AI frameworks, proxies, and agent deployments to find attackers building AI-native playbooks.
Research & Knowledge
Cyera
Cyera broke down the OpenAI/Hugging Face agentic attack, where an AI agent went from a stuck evaluation task to cluster admin access in under 13 hours.
Research & Knowledge
Cequence Security