AI Security Articles

AgentSeal is the open-source security toolkit that scans AI agents for dangerous skills, poisoned MCP configs, and prompt injection vulnerabilities. With 225+ attack probes, real-time monitoring, and offline-first protection for 28 agents, it's the defense layer AI development desperately needs.

Cisco Skill Scanner is the first open-source security tool purpose-built for AI agent skills. Detect prompt injection, data exfiltration, and malicious code in OpenAI Codex, Cursor, and Claude Code skills using multi-engine static, behavioral, and LLM analysis.

Clawdstrike is a fail-closed policy engine and cryptographic attestation runtime for AI agent systems. Learn how this open-source EDR alternative stops shadow agents, signs every decision with Ed25519 proofs, and provides formally verified security enforcement at the tool boundary.

Crust is an open-source AI agent security gateway that intercepts and blocks dangerous tool calls before execution. With 42 built-in rules, 51 DLP patterns, and zero-config setup, it protects your secrets from prompt injection and exfiltration attacks.

DefenseClaw by Cisco AI Defense is an open-source security governance layer that scans and blocks dangerous AI agent actions before they execute. Learn installation, real code examples, and why traditional security tools fail against agentic AI threats.

MEDUSA is the AI-first security scanner with 9,600+ detection rules for repo poisoning, MCP attacks, and LLM vulnerabilities. Learn how to scan any GitHub repo in seconds and protect your AI supply chain.

Discover how 0DIN AI Scanner's 179 community probes built on NVIDIA garak help organizations find vulnerabilities before deployment. Complete setup guide, real code examples, and comparison with alternatives included.

Discover KnAIght, the open-source AI prompt obfuscation tool from WKL-Sec that's helping security researchers bypass detection systems. Complete guide with installation, code examples, and advanced techniques for 2025.

Discover nono, the kernel-isolated sandbox from Sigstore's creator. Zero-latency capability-based security for AI agents with credential injection, snapshots, and cross-platform support. Stop hoping your agents won't be compromised—guarantee they can't be.

G0DM0D3 is a fully open-source, privacy-respecting multi-model chat interface for AI red teaming. Built as a single HTML file with 55+ models, parallel jailbreak racing, and systematic perturbation engines. Deploy anywhere in seconds.

ClawGuard is a revolutionary security dashboard for AI agents with real-time monitoring, intelligent risk analysis, and an emergency kill switch. This comprehensive guide covers installation, code examples, and advanced usage for developers deploying autonomous systems.

Learn how to protect your sensitive information when using AI tools. This comprehensive guide reveals why data masking is critical, real-world cases of privacy breaches, step-by-step safety protocols, top tools including Pasteguard, and industry-specific use cases. Includes a free infographic checklist.