Game-Theoretic AI for Robust Course of Action (COA) Generation and Wargaming

# Q&A Summary of Changes **New Q&A Added (Q1):** Substantially expanded guidance on game-theoretic foundations, covering rationality assumptions, Nash equilibrium vs. mixed strategies, human irrationality, doctrine adherence vs. optimality, objective function definition, and human-AI integration workflows. Key clarification: relaxing rationality assumptions is acceptable; mixed strategies are permissible if explainable; doctrine deviation is appropriate for this effort. **No other substantive changes:** Q2–Q4 remain identical to previous version with "proposers should do what seems most feasible or optimal" responses.

# Q&A Changes Summary **New Question Added (Q4):** Clarifies TABA (Technical Assistance and Business Acceleration) eligibility under Direct-to-Phase II. Answer confirms SCO does not provide TABA support or preferred vendor lists, but directs proposers to sbir.gov/lab2market for resources. **Clarified Answer (Q5 - formerly Q4):** Adds critical security requirements: Classified work requires active Top Secret clearance with SCI eligibility and facility clearance (minimum Top Secret) **at time of award**. Clearances cannot be sponsored during performance due to schedule constraints.

**Q&A Update Summary:** Added 15 new questions (Q1-Q3 with 15 items total) covering: adversary modeling approaches, robustness testing priorities, data rights/classification assumptions, Phase II success criteria, stakeholder alignment, imperfect information handling, M&S environment specifics, interpretability requirements, computational constraints, response-time thresholds, prototype maturity standards, mission scale, COA detail levels, performance metrics, and game-theoretic approaches. **Key change:** Government response to all new questions: "proposers should do what seems most feasible or optimal to their business and technology" — indicating maximum flexibility on technical approach rather than prescriptive requirements. Original Q4 (security clearances, M&S environment, red teams, scoring functions) retained with identical answers reaffirming: Top Secret/SCI clearance required at award (not sponsorable), M&S environment designated post-Phase I, proposers provide SMEs for testing, government determines scenario detail level.

**Summary of Q&A Changes:** One answer was updated regarding facility clearance requirements. The facility clearance level was raised from "minimum Secret level" with "Top Secret may be required" to "minimum Top Secret level is expected." Additionally, language stating "Top Secret facility clearance may be required depending on the designated modeling & simulation environment" was removed, making the Top Secret facility clearance requirement definitive rather than conditional.

This Q&A clarifies Phase I and II requirements for an AI modeling/simulation solicitation (OSW26BZ02-DV004), including that the government will select the M&S environment post-Phase I, offerors must provide subject matter experts for testing, and selectees require active Top Secret/SCI clearance and minimum Secret facility clearance at award.

Status changed from Pre-Release to Open

New opportunity: Game-Theoretic AI for Robust Course of Action (COA) Generation and Wargaming