RSAC 2026 Offers a Glimpse into the Future of Offensive Security

RSA Conference is infamously noisy, but noise does not always equal empty hype, and the buzz around AI penetration testing was certainly real. Security leaders are rearing for the shift from point-in-time security testing, to continuous and risk-aware penetration testing. But it’s the boom days of autonomous security testing, which means lots of new acronyms, lots of new tech, and lots of difficulty distinguishing one vendor from another.

Here’s our 4-min-read breakdown of the biggest offensive security takeaways from RSAC:

How Continuous Pentesting Crossed from Concept to Competitive Market

Just weeks before RSAC, Gartner released their guide to continuous offensive security testing featuring Novee (read more about how Novee meets their model here). And 34% of survey respondents on Latio’s 2026 AppSec Report listed “AI Pentesting” as the AI feature they’re most excited about – the most of any category. In short, autonomous offensive security is top of mind for security leaders.

If you’re evaluating this space, the practical question is no longer whether to adopt AI-driven offensive testing, but how to distinguish platforms that genuinely replicate attacker behavior from those that automate existing scanning workflows. 

Testing AI Applications Is the Unsolved Problem Everyone Is Talking About

AI was everywhere at RSAC: building it, using it, guarding it, debating it. Another challenge is figuring out how to run successful, actionable security tests on it.

The shift to AI-powered products has created an offensive security blind spot that the industry is only beginning to address. Enterprises are shipping AI-powered applications, like chatbots, internal copilots, and autonomous agents, faster than security teams can evaluate their risk; LLM functionality increases attack surface by adding statefulness and trust complexity to application environments.

Traditional pentesting tools were built for web applications, networks, and APIs. They have no framework for how an LLM interprets prompts, chains reasoning steps, or interacts with external systems. And human pentesters with expertise testing AI-enabled applications are a rare breed, and unable to scale their efforts. That gap leaves organizations exposed to a class of vulnerabilities (prompt injection, jailbreaks, agent manipulation, silent data exfiltration) that scanners and manual testers weren’t designed to find.

ICYMI: Live Demos of What Your Attackers Already Know 

RSAC was Novee’s first major conference debut since launching out of stealth, and we came ready. Here’s a quick summary of our week:

• Launched AI Red Teaming for LLM Applications. Our new capability for autonomously testing AI-enabled systems, including chatbots, copilots, autonomous agents, and LLM-powered workflows, for the full range of AI-specific attack techniques.
• Won two Global InfoSec Awards:
  • Most Advanced AI Security Solution
  • Transformational Pentesting
• Live Demos at our Custom-Built Booth. Starting from a domain name and showing real attack chains, validated findings, and personalized remediation. Plus, no theme park lines or carnival games required to get a demo.
• Released the Definitive Buyer’s Guide to AI Penetration Testing. The 8 questions to ask any AI pentesting vendor before you commit.
• Featured on CyberRisk TV. CEO Ido Geffen on what separates true AI penetration testing from vulnerability scanning

If you didn’t make it to the booth, book a demo and we’ll show you how – with nothing but your domain name – Novee maps the full attack surface, executes multi-step attack chains, validates exploitability, and delivers actionable remediation guidance. All continuous, with no humans in the loop.

Top AI Pentesting and AI OffSec Sessions at RSAC 2026

Ground your knowledge of the market with these education-focused talks from RSAC (recordings available to view with a conference log-in):

“How to Red Team a Frontier AI Model.” Adversarial testing of AI systems in production is no longer optional. Methodology, case studies, practical takeaways.

“From Prompt to Pwn: Exploiting Web Apps with LLM & OWASP Techniques.” How traditional OWASP exploitation can work in tandem with AI vulnerabilities like prompt injection, agent manipulation, and LLM trust boundary abuse. Adding LLM functionality increases statefulness, trust complexity, and attack surface simultaneously.

“Beyond Red Teaming: Why AI Security Needs a Bigger Playbook.” AI systems behave dynamically, and static validation models will struggle to keep up.

How to Actually Evaluate your AI Pentester

With so many vendors entering the AI pentesting space simultaneously, the practical challenge for security leaders is differentiation. 

The market now includes everything from legacy scanner vendors adding an “AI” label to purpose-built autonomous systems trained on real attacker tradecraft. The capability gap between them is significant.

RSAC is an invaluable opportunity not just for innovators and engineers to show off, but for solution-seekers to get in front of the best tools for their needs. Finding the right fit is about:

(a) understanding where the AI pentesting market is headed, 

(b) asking the right questions, and

(c) seeing the tech in action. Book a demo with us to see how Novee does it.