Resources

Research

We Tested GPT-5.5 for Offensive Security. It's Good.

OpenAI dropped GPT-5.5. We ran it through our platform — here's where it wins for offensive security, and where Opus still leads.

Technical

Early-Stage AI Hacking Agent Secures Top 8% on Global Leaderboard

Our single, general-context AI pentesting agent went head-to-head at the HackTheBox Neurogrid CTF — pure autonomous reasoning in a live, adversarial environment.

Open Source

PentestShield: Private AI-Assisted Pentesting with an MCP Proxy

PentestShield is an open-source MCP proxy that redacts sensitive data before your LLM sees it — keeping client data private during AI-assisted pentesting.

CVE

CVE-2025-68613 — Critical RCE in n8n Workflow Automation

A critical (CVSS 9.9) RCE in n8n lets authenticated users execute arbitrary code via unsafe expression evaluation — with 100,000+ instances potentially exposed.

Research

How Strike7 Uses Agent Skills for Autonomous Attack Execution

How Strike7 packages offensive capabilities as modular, self-describing Agent Skills the agent selects by context — for adaptive, governed autonomous pentesting.

Technical

Fully Patched, Still Broken: Strike7 vs GOAD Hard Mode

We handed Strike7 a single /24 and the words “run a network pentest” — no hints, fully patched GOAD hard mode — and it reasoned its way to full forest-root compromise.

Company Update

Introducing Strike7: Autonomous Offensive Security at Attacker Speed

Strike7 is an autonomous AI platform that brings agentic intelligence to offensive security — amplifying pentesters, red teams, MSSPs, and security engineers.

Research

From Human Hackers to Autonomous AI Attackers: What GTG-1002 Means for Security Leaders

Anthropic's GTG-1002 was the first confirmed AI-orchestrated espionage campaign. Here's why point-in-time pentesting can't keep up with autonomous attackers.