How Do You Safeguard AI When Development Outpaces Security? With Ante Gojsalić - SplxAI

Generative AI is moving faster than our defences — can we catch up?

In this episode of Razorwire, host James Rees (aka Jim) speaks with Ante Gojsalić, CTO and co-founder of SplxAI, to dissect the growing risks, complexities, and opportunities in securing AI systems before they outpace our ability to protect them.

From the global AI arms race to the rise of agentic decision-making, this episode cuts through the hype to explore the real vulnerabilities in generative AI. You’ll learn why securing AI is not just a DevSecOps problem, how attackers are already exploiting new behaviours, and what CISOs, engineers, and business leaders can do right now to future-proof their approach.

🧠 “With AI, which is non-deterministic and still changeable day by day… it’s just unstable.” – Ante Gojsalić

🎯 Key Talking Points:

• Why traditional security models fail in the world of generative AI
• Hidden risks of autonomous AI agents handling data and making decisions
• The collision between innovation speed and long-term security priorities
• How to embed continuous testing and secure-by-design thinking from day one

🎧 Listen now on your favourite platform: (https://razorwire.captivate.fm/listen)

⸻

🔍 In This Episode:

• The Rise of Generative AI – What it is, how it works, and where it’s going
• Global AI Development – East vs West approaches and what it means for you
• Agentic Risks – Why AI that acts like humans creates security blind spots
• Continuous Pen Testing – Why one-time scans won’t cut it anymore
• Business vs Security – How to keep pressure from stakeholders from undermining AI safety
• Social Engineering 2.0 – AI-powered phishing and manipulation is already here
• Governance & Regulation – Where compliance frameworks fall short
• Securing AI from the Start – Actionable tips for CTOs, CISOs, and developers
• Future-Proofing AI Security – Planning for what happens after AI becomes embedded everywhere

⸻

💡 Mentioned in This Episode:

• SplxAI & Probe (https://splx.ai)
• OWASP, DevSecOps, PCI, DORA
• OpenAI, DeepSeek, Mitra, Ray Kurzweil
• The Cyber Sentinels Handbook by James Rees
• Razor’s Edge Continuous Pen Testing Service

⸻

🎙️ About Your Host
Hi, I’m James Rees, host of Razorwire and founder of Razorthorn Security. With over 25 years in cybersecurity, I’ve seen firsthand how the threat landscape evolves — and this podcast exists to help you stay ahead of it.

Each episode brings you real conversations with security professionals who are on the frontlines of protecting data, systems, and people. From seasoned leaders to up-and-coming voices, we explore the ideas and innovations shaping the future of cyber.

⸻

📌 Subscribe & Connect
🌐 Website: (https://www.razorthorn.com)
📧 Email: podcast@razorthorn.com
📍 LinkedIn: (https://www.linkedin.com/company/razorthorn-security)
📍 YouTube: (https://www.youtube.com/@RazorthornSecurity)
📍 Twitter/X: @RazorThornLTD