AI Agent Security Research

Analysis, threat intelligence, and strategic perspectives on securing the emerging agent-to-agent ecosystem. Written for security leaders and investors.

ProductMar 2026~6 min read

Oktsec Plugin for OpenClaw: Scanning Every Message and Tool Call

Oktsec now runs as a native plugin inside OpenClaw. 217 detection rules scanning every message and tool call before execution, independent from the model. Session management, egress policies, and OpenClaw-specific security rules.

EngineeringMar 2026~12 min read

oktsec v0.10.0: Delegation Chains, LLM Escalation, Scan Profiles

Delegation chains with Ed25519 signed tokens, LLM verdict escalation, scan profiles for tool-scoped sensitivity, session traces, ephemeral keys, and a Bubbletea terminal UI.

Industry AnalysisMar 2026~14 min read

Before RSA 2026: Three Agent Security Frameworks Every CISO Needs

OWASP Top 10 for Agentic Applications, CoSAI 40-threat taxonomy, and MAESTRO 7-layer model converge two weeks before RSA 2026. A deep analysis for CISOs.

EcosystemMar 2026~12 min read

Aguara v0.5.0: 173 Detection Rules and Confidence Scoring Strengthen the Oktsec Stack

20 new detection rules and a confidence scoring system. For enterprises running Oktsec, this means smarter content scanning, fewer false positives, and threshold-based CI/CD gating across 177 total detection checks.

ReleaseMar 2026~7 min read

AI Agent Security: Checklist and Guide

Free checklist (28 controls, 3 tiers) and 55-page guide covering the AI agent threat landscape. Based on 58,000+ scanned skills and every documented incident through March 2026.

SecurityMar 2026~9 min read

AI Agents Don't Understand Secrets: 23.8M Credentials Leaked on GitHub in 2024

Copilot-using repos show a 40% higher secret leak rate. Five vulnerability paths, why pre-commit hooks aren't enough, and defense-in-depth with Aguara and Oktsec.

Industry AnalysisFeb 2026~14 min read

The CSA Agentic Trust Framework: Zero Trust Governance for AI Agents

CSA published the Agentic Trust Framework applying Zero Trust to AI agents with Intern-to-Principal maturity levels. Every trust element maps to a shipping Oktsec feature.

SecurityFeb 2026~12 min read

Academic Research Validates the Zero-Trust Runtime Architecture Oktsec Implements

Researchers at Northeastern, NYU, UCSD, and UIUC formalize the agentic AI threat model. Every recommendation maps to a shipping Oktsec feature.

EngineeringFeb 2026~14 min read

Oktsec v0.6.0: MCP Gateway, Security Hardening & Coordinated Stack Release

MCP Gateway mode, 21 SSRF CIDR blocks, 15 credential redaction patterns, 17 MCP client discovery, and official Linux Foundation SDK migration across the entire stack.

ProductFeb 2026~10 min read

MCP Gateway: A Security Layer for Every MCP Server

Oktsec now acts as a security gateway for any MCP server. Per-agent identity, tool-level policies, content scanning, and audit trails without changing backend code.

SecurityFeb 2026~9 min read

Kali Linux Just Proved That AI Agents Are Offensive Security Tools

Kali Linux officially integrates Claude Desktop to control pentesting tools via MCP. When agents can run nmap and metasploit through natural language, policy enforcement and identity verification become requirements.

Industry AnalysisFeb 2026~10 min read

NIST's AI Agent Identity Paper Validates the Oktsec Thesis

NIST NCCoE asks 6 questions about AI agent identity, authentication, and authorization. Oktsec already implements answers to all of them.

EcosystemFeb 2026~8 min read

From Open-Source Scanner to Security Platform in 14 Days

How Aguara's open-source scanner, observatory, and MCP server created a compounding data advantage. 42,655 skills, 7 registries, 188 detection rules. The platform strategy behind Oktsec.

SecurityFeb 2026~10 min read

Static Analysis + Runtime Isolation: A Defense-in-Depth Framework for AI Agents

Why neither Docker Sandboxes nor static scanning alone is enough. A practical framework combining pre-runtime analysis with runtime isolation for enterprise AI agent deployments.

SecurityFeb 2026~11 min read

The OpenClaw Security Crisis: What Enterprise Teams Need to Know

CVE-2026-25253, 76 confirmed malicious skills, 42,665 exposed instances. A concise briefing on the risks and mitigations for teams evaluating AI coding agents.

MarketJan 2026~9 min read

Why Agent-to-Agent Security is the Next Enterprise Requirement

Gartner projects 33% of enterprise software will include agentic AI by 2028. As agents communicate with agents, security shifts from perimeter defense to message-level verification.