Risk-Aware Regenerative AI-based Multimodal Visual-Tactical (ISRT) (Observant-AI) – Monitor, Understand, Alert, and Assist

The DON26BZ01-NV023 project, "Observant-AI," seeks to develop risk-aware, regenerative AI-based computing methods to enhance naval situational awareness and decision support. Motivated by three naval challenges, the project aims to create a distributed system of AI agents that self-organize, share insights, and form collaborative teams. These agents will process and fuse multidomain anomalous events for real-time visual-tactical understanding, monitoring, alerts, and operational risk assessment. The Observant-AI will provide natural language explanations for human-AI interactions and assist with courses of action and countermeasures in real-time. Key objectives include "LIVE" machine self-teaching, contextual machine exploitation, contextual networking for all-source intelligence, and proactive AI-assisted targeteer and decision support. The technology will address vulnerabilities in current Naval ISRT operations, such as human biases and limitations of generative AI when encountering "Unknown-Unknown" events. Phase I involves technical feasibility studies, distinguishing the approach from current systems, and motivating the design with challenge problems. Phase II focuses on proof-of-concept and prototype development, with a probable classification of work. Phase III aims for TRL-7 advancement and integration into naval command and control systems, with dual-use applications in civilian law enforcement and security services. The project emphasizes stringent CMMC Level 2 requirements and ITAR restrictions, requiring disclosure of any foreign national involvement.