Full Segment Notes
Adding AI agents to an unprepared SOC doesn't make it smarter; it just makes it "faster at being wrong." Georges Bossert challenges the industry hype to explain why true autonomy relies on reliable context and structured runbooks, not just prompts. He will discuss how to build the necessary foundations to automate rapidly without losing control.Segment Resources:AI SOC Whitepaper: https://go.sekoia.io/en-ai-soc.html?utmsource=scworld&utmmedium=referral&utm_campaign=RSAC2026Sekoia AI-Guided SOC Demo: https://www.youtube.com/watch?v=1qkXeseSvE0Sekoia TDR threat investigations: https://blog.sekoia.io/?utmsource=scworld&utmmedium=referral&utm_campaign=RSAC2026This segment is sponsored by The Sekoia.io. Visit https://securityweekly.com/sekoiarsac to discover their AI SOC Platform!Read the interview summary from SC Media here: Sekoia.io’s Georges Bossert on avoiding making SOCs faster at being wrong
Key Moments
- 0:00 - Intro – Agentic AI in Cybersecurity (RSAC 2026)
- 0:30 - The Growing Threat of Agentic AI
- 01:00 - AI in SOC: Opportunity vs Risk
- 01:26 - Why AI Adoption Starts with Data (Not LLMs)
- 02:24 - “Garbage In, Garbage Out” in AI Security
- 02:49 - Why Generic LLMs Fail in SOC Environments
- 03:35 - The Danger of “Failing Faster” with AI
- 03:54 - AI Agents = Junior Analysts (Key Analogy)
- 04:55 - Why Context Is Critical for AI Security
- 05:37 - AI vs Humans: The Problem of False Confidence
- 06:10 - When AI Lies (And Why It’s Dangerous)
- 07:06 - Chatbots vs Real AI Reasoning Systems
- 07:39 - How Runbooks Power AI Security Automation
- 09:29 - Feeding Context: Data, Users, Networks & Threats
- 09:35 - Can AI Agents Actually Improve Over Time?
- 10:40 - AI Reasoning Playbooks Explained
- 11:41 - Why Chain of Thought Matters in Cybersecurity
- 12:34 - The Rise of AI-Powered Threat Detection
- 13:22 - Biggest Risk: Rushing AI in the SOC
- 14:45 - When AI Slows You Down Instead of Helping
- 15:12 - The 3 Pillars of Effective AI Security Operations
- 16:57 - Final Thoughts & Key Takeaways
Guest
Georges Bossert is the Co-founder and Chief Technology and Product Officer of Sekoia.io. An engineer by training and PhD graduate from CentraleSupélec, his research focused on applying machine learning and grammatical inference to cybersecurity.
With over 15 years of experience, he has worked across the full spectrum of cybersecurity — from reverse engineering to leading technology and product strategy.
A former reservist in the French Army’s cyber operations division, Georges now serves as a board member and lecturer at the University of Rennes. He is passionate about innovation, mentoring, and building resilient teams. He was named *Innovator of the Year 2024* by *Le Point* magazine and a 2025 *Cyberscoop 50* finalist.
