Controlling Advanced AI: The Gap Between Theory and Reality

As governments struggle to establish clear AI regulations, organizations are left navigating the frontier on their own. Meanwhile, blanket restrictions are failing. When access to proprietary models is blocked, demand quickly shifts to open-weight alternatives. This forces companies to establish their own internal governance and adoption frameworks, resulting in a familiar business challenge, amplified by AI: how do you innovate to stay competitive while maintaining the security, oversight, and control necessary to protect the enterprise?

Recent events offer sharp illustrations of this challenge. Export controls on advanced models shift demand rather than eliminate capability, but they're one policy lever among many. Governments globally are grappling with how to govern AI in ways that protect critical infrastructure without stifling innovation or sovereign capability. The EU is codifying frameworks. Others are still determining what governance means. The real gap is between the theoretical need for control and the practical challenge of defining what should be controlled, how, and by whom.

Effective governance demands moving beyond the assumption that restriction alone works. It requires rethinking security architecture, organizational readiness, and how institutions prepare for increasingly distributed technology.

Join the discussion to learn more about:

• The disconnect between how we attempt to control frontier AI and how it actually plays out in practice
• What recent real-world incidents reveal about the limits of current governance approaches
• Identifying the genuine dual-use risks that matter most
• Building organizational resilience in an era of advanced AI
• Moving beyond restriction to fundamentals: architecture, governance, and preparation