Adversaries don’t wait for your next assessment window. They probe continuously, adapt in real time, and exploit the gap between when your Zero Trust architecture was approved and when it was last tested. For most DAF programs, that gap is measured in months, sometimes years.

The 2026 DAFITC theme demands speed: out-pacing adversary development cycles and ensuring resilience across contested environments. The Department is actively shifting from compliance-centric ATOs to data-driven, automated security, moving away from point-in-time snapshots toward continuous authorization and real-time assurance. Autonomous penetration testing sits at the exact intersection of those two imperatives. It is AI-native capability that continuously validates whether Zero Trust architecture holds under real adversarial conditions, not just on paper and not just once.

This presentation introduces ARES, an AI-native autonomous penetration testing agent that replicates the full cognitive loop of an expert red team operator without constant human direction. Attendees will walk through how ARES chains reconnaissance, lateral movement, and exploitation autonomously across enterprise and OT/ICS environments; how its persistent cross-session memory accumulates adversarial intelligence over time rather than resetting after each engagement; and how its reflection engine continuously reassesses and pivots strategy mid-operation. The presentation will demonstrate how ARES maps directly to DoD Zero Trust pillars, with primary coverage across Network/Environment, Applications and Workloads, and Devices, and how it integrates into existing DAF security stacks as a continuous validation layer rather than a replacement for existing tooling.

Attendees will leave with a concrete framework for thinking about Zero Trust validation as a continuous operational discipline rather than a compliance milestone. For program managers, that means understanding how autonomous adversarial testing integrates into cATO workflows and reduces the overhead of recurring manual assessments, turning ZT assurance into a reportable, auditable data stream rather than a periodic event. For cyber operators, it means a practical look at how AI-driven red teaming augments human teams by handling repetitive reconnaissance and lateral movement work so operators can focus on decisions that require human judgment. Both audiences will leave with a clear picture of what continuous autonomous validation looks like in a DAF-relevant environment.

What separates this presentation from broader Zero Trust and automation discussions is operational specificity. ARES is not a scanner, a compliance tool, or a theoretical framework. It is an autonomous agent that thinks like an adversary, accumulates intelligence across engagements, and self-corrects without human re-initiation. Critically, it operates fully air-gapped with no cloud reachback required, making it viable in the classified and disconnected environments where DAF programs actually operate. And unlike point-in-time red team engagements that require scarce human expertise, ARES runs continuously, meaning the adversarial pressure on your Zero Trust architecture never stops, even when your red team does.