AI Advances Demand An Inside-Out Approach For Security And Resilience

Chris Konrad is VP of Global Cyber at WWT, leading its multi-billion-dollar cybersecurity business for enterprise and public-sector clients.

​Your biggest emerging security threat might be lurking inside your network today. It might have direct access to some of your most critical systems. It might have permission to make decisions like a senior executive and implement changes that could significantly impact your business. And it exists because you created it and empowered it.

​Developments in frontier AI over just the past several weeks have forced us to expand long-standing traditional security practices. Gone are the days when we could focus most of our energy on locking down systems and models from external threats, even though those external risks continue to grow by the day. We must now also govern the autonomous systems operating inside the enterprise.

​It’s a daunting proposition, but there is a significant silver lining: with the right safeguards and proper human oversight, those AI tools can shift from potential risks to powerful allies in maintaining security in an environment that can change in the blink of an eye.​

​Enterprises can effectively make this pivot by focusing on 3 key steps:​

​1. Assess And Govern The Scope Of AI’s Access And Authority​

​For decades, network security was built on answers to three key questions: What’s on your network? Is it supposed to be there? What is it doing?​

​This approach was and continues to be highly effective in identifying, preventing and mitigating external threats. But today, we must apply the same rigor to our own AI agents. Can you see where they are within your systems and what they’re doing there? Just as importantly, how much authority do they have to make changes without human oversight?​

​Each AI agent should have a clear, differentiated identity, narrowly scoped permissions, observable activities and a clear path to revoking authority should an issue arise.​