
You’re about to see the world’s first show & tell from a hackbot. Enjoy!

Autonomous Ethical Hacking

You’re about to see the world’s first show & tell from a hackbot. Enjoy!

🚨 Running NGINX? These two new critical CVEs deserve your attention. CVE-2026-123633 and CVE-2026-123007, both rated 9.2 (Critical) under CVSS 4.0, are memory corruption vulnerabilities that could allow remote code execution if HTTP/2 is enabled. Our security research team is already investigating these vulnerabilities and exploring a dedicated testing module. In the meantime, with Ethiack’s Exposure Check, you can check your domain for publicly exposed services, technologies (including NGINX), and recent CVEs that may affect them. Start by checking your domain’s exposure at exposure.ethiack.com/

Java4Shell (CVE-2026-114938) has put Java exposure back in the spotlight 🔎 Following yesterday’s disclosure, this new vulnerability in OpenJDK’s Base64 decoder could allow remote code execution under specific conditions. With our free Exposure Check, you can check your domain for publicly exposed services, technologies (including Java), and recent CVEs that may affect them. Could your domain be exposed? Find out at exposure.ethiack.com


🚨 CVE-2026-114938 (Java4Shell): OpenJDK 17.0.6+ may let attackers run code via crafted Base64. Fixed in Oracle 17.0.21.1. No credentials needed, AVX-512 CPUs only, and no exploitation in the wild reported as of 17 Nov. ethiack.com/info-hub/blog/ja…





1. Manual pentesting: high depth, low coverage, narrow scope, annual cadence. 2. Legacy scanning: low depth, high coverage, signature-based, no chaining, no proof. 3. PTaaS/hybrid: mid-depth, mid-coverage, splits the difference. 4. Agentic AI pentesting: chains vulnerabilities, generates payloads dynamically, across full surface, continuously, narrowing the gap for known exploit classes. Human depth still matters most for novel business-logic chains. Retail, telco, and fintech programmes increasingly need both quadrants, not one.

Manual vs. Automated Pentesting Is the wrong debate Every article on this ends the same way: manual for depth, automated for scale, use both. True in 2018. Incomplete in 2026. "Automated" now splits into legacy scanning that flags potential issues, and agentic AI pentesting that attempts to prove what an attacker can exploit. Here's the coverage-vs-depth quadrant (see image) to reflect this.

Anova Solutions (@AnovaSolutions) needed to know their attack surface in real time, not once a year. Here's how continuous exposure validation enabled this for their security team

Want to test your own pentesting agent? EthiBench is open source: github.com/ethiack/ethibench


