Research & Analysis

Blog

Security research, AI agent benchmarks, vulnerability analysis, and best practices from the OpenA2A team.

Introducing OASB: The Security Benchmark for AI Agents

OASB defines the first comprehensive security benchmark for AI agents — 46 controls across 10 categories with 3 maturity levels.

OpenA2A Team|February 10, 2026
oasbbenchmarksecurity

OpenClaw Merges Built-In Skill Security Scanner

PR #9806 merged 1,721 lines into OpenClaw adding a code safety scanner that detects malicious patterns in skills on install and update.

OpenA2A Team|February 6, 2026
openclawsecurityopen-source

CVE-2026-25253 Now Has a Scanner: Detecting the OpenClaw WebSocket RCE

HackMyAgent v0.4.0 ships the first automated detection for CVE-2026-25253 (CVSS 8.8), expanded ClawHavoc IOCs, and 11 new security checks.

OpenA2A Team|February 5, 2026
cve-2026-25253openclawclawhavochackmyagent

The State of AI Agent Security: 97,000 Hosts, 1,190 Exposed Configs

We scanned 97,013 internet-facing hosts for AI agent vulnerabilities. 14.4% had confirmed security issues.

OpenA2A Team|February 5, 2026
security-researchai-agentsmcp

I Broke My AI Agent in 5 Minutes (And You Should Too)

HackMyAgent is a security toolkit for AI agents with 4 modes: Attack, Secure, Benchmark, and Scan.

OpenA2A Team|February 4, 2026
hackmyagentsecuritytutorial

Why Your NHI Strategy Doesn't Cover AI Agents

Traditional NHI platforms manage service accounts and API keys. AI agents are a fundamentally different class.

Abdel Fane|February 2, 2026
nhigovernancesecurity

341 Malicious Skills and a 1-Click RCE: Scanning OpenClaw for ClawHavoc

The ClawHavoc campaign planted 341 malicious skills on ClawHub. We built a scanner to detect it.

OpenA2A Team|January 31, 2026
openclawsupply-chainhackmyagent

The OWASP Agentic Top 10 and What It Means for NHI Governance

How each OWASP Agentic risk maps to NHI governance capabilities.

Abdel Fane|January 26, 2026
owaspagentic-aicompliance

The ServiceNow AI Vulnerability: What Went Wrong

ServiceNow disclosed the most severe AI-driven vulnerability to date.

Abdel Fane|January 15, 2026
vulnerabilitycase-studyai-security

Introducing AIM: Open Source Security for AI Agents

AIM provides cryptographic identity, MCP attestation, trust scoring, and audit logging.

OpenA2A Team|December 16, 2025
launchaimopen-source

One Line of Code to Secure Your AI Agents

CVE-2025-32711 (EchoLeak) affected Microsoft Copilot. Learn how to secure your AI agents.

Abdel Fane|November 7, 2025
securityecholeakmcp