AI Agents vs. AI Agents: The Future of Security — Monzy Merza on Secure and Simple

Cybersecurity is entering an era where the fight isn't human versus human. It's agent versus agent.

In this episode of the Secure and Simple podcast, host Dejan Kosutic from Advisera sits down with Monzy Merza, co-founder and CEO of Crogl, to unpack what it means when attackers can task AI agents to run sophisticated, low-cost campaigns without any human approval loop, and what defenders need to do to keep pace.

What We Cover

• Why cybersecurity is shifting to an "agent versus agent" world
• The three pillars of security operations (preparation, alert investigation, and response) and how AI is changing each
• How AI SOC agents connect to multiple data sources, enrich alerts, run MITRE kill chain analysis, and produce full investigation reports automatically
• When humans must stay in the loop, and when it's safe to hand off to AI
• How organizations build trust in AI through phased adoption with measurable use cases
• Why security roles may shift from analysts toward more security engineers
• What governance looks like: flexible integrations, model choice, and transparency

Episode Highlights

• Attacker economics: AI lets adversaries run fast, sophisticated campaigns at near-zero cost and without human approvals, compressing the window defenders have to respond
• The three pillars: Preparation and tooling, alert investigation, and response each have distinct AI leverage points
• Autonomous investigation: AI SOC agents that automatically pull data from multiple sources, enrich context, map to the MITRE ATT&CK kill chain, and deliver a completed investigation report
• Human-in-the-loop: High-impact response decisions still require human judgment; the goal is to put humans in the right decisions, not every decision
• Phased trust: Organizations adopt AI in stages, starting with measurable, low-risk use cases and expanding as confidence grows
• Role evolution: The analyst role may shift toward security engineering as AI handles more of the routine investigation work
• Governance requirements: Flexibility in data integrations, choice of underlying models, and full transparency in how AI reaches its conclusions

Also available on Spotify and Apple Podcasts.