DEF CON 34 – Main – Defender’s Dilemma Releasing Dual-Use AI Models as Capabilities Climb – Panel

Defenders and attackers are locked in a renewed race where AI has enabled new defenses and new attacks. AI models are gaining capability at a remarkable pace, and dual-use systems, powerful for defenders and adversaries alike in domains like cybersecurity, bring that trajectory into sharp focus. As these models grow stronger, developers face a practical question: How can defenders and labs each play a part in ensuring that defenders win?

This panel takes up emerging best practices for applying and releasing dual-use models safely, from how teams evaluate a model’s capabilities before release to the release-strategy choices that follow. Panelists will weigh approaches and how those decisions shift as capabilities advance. Looking ahead, the conversation will consider what the next wave of increasingly capable models means for safely deploying AI models and how developers can prepare now for the exponential to come.DEFCONConferenceRead More