DAO Governance
BASTION implements multi-DAO governance aligned to military command echelons, with a full proposal lifecycle, coalition support, and AI agent integration under strict safety constraints.
Three-Tier DAO Structure
| Tier | Scope | Example |
|---|---|---|
| Strategic | Theater/national-level policy and guidance | INDOPACOM Strategy DAO |
| Operational | Campaign and major operation decisions | JTF Planning DAO |
| Tactical | Execution-level task approval | Battalion Ops DAO |
Hierarchical Relationships
DAOs form parent-child hierarchies. Child DAOs inherit membership and policy from their parent, ensuring strategic guidance flows down while tactical feedback flows up. A Strategic DAO can spawn multiple Operational DAOs, each of which can spawn Tactical DAOs.
Proposal Lifecycle
- Draft -- Author creates proposal with required metadata
- Screen -- AI proposal-screener checks completeness and policy compliance
- Context -- AI context-analyzer enriches with operational context
- Feasibility -- AI feasibility-assessor evaluates resource and risk factors
- Review -- Human reviewers examine AI-enriched proposal
- Vote -- Members vote under the configured voting scheme
- Execute / Reject -- Approved proposals trigger execution; rejected proposals are archived with rationale
Proposal Kinds
Beyond standard proposals, BASTION supports domain-specific types:
| Kind | Purpose |
|---|---|
| PhaseTransition | Advance the operation to the next phase |
| AssumptionAcceptance | Formally accept or reject a planning assumption |
| ProductApproval | Approve a planning product (e.g., OPORD annex) |
| RedTeamGate | Require red team review before proceeding |
| CommanderGuidance | Issue or update commander's guidance |
| StrikeAuthorization | Authorize kinetic or cyber strike -- always human, 100% approval threshold |
Voting Engine
Each DAO configures one of three voting schemes:
- Token-weighted -- Votes proportional to governance token holdings
- Role-weighted -- Votes weighted by the voter's staff role and echelon
- Equal -- One member, one vote
The voting scheme is set at DAO creation and can be changed only by a governance proposal within the parent DAO.
Coalition Proposals
Multi-party operations (Five Eyes, NATO, ad hoc coalitions) use coalition proposals that require approval from each participating DAO. A coalition proposal is not enacted until every designated party reaches its own approval threshold independently.
Five-Tier Authority Delegation Model
BASTION implements authority delegation through five echelon-based tiers, each enforced by smart contract:
| Tier | Scope | Example | Engagement Authority |
|---|---|---|---|
| Tier 1 | Individual | Single operator within bounded scope | None — administrative only |
| Tier 2 | Team | Small unit coordination | None — coordination only |
| Tier 3 | Organizational | Commander-level authority | Minimum for autonomous engagement |
| Tier 4 | National DAO | National resource commitments | Strategic resource decisions |
| Tier 5 | Coalition Strategic | Full coalition consensus | 100% approval for strike authorization |
Delegations are dynamic and time-bounded. A commander can grant expanded delegation for a limited mission window (e.g., "Tier 3 engagement authority for four hours in grid 12S"). The delegation is recorded on the NEAR blockchain with its expiration time and expires automatically. No one has to remember to revoke it.
Interaction-Based Authority Levels (MDMP)
In addition to echelon-based tiers, each MDMP planning activity is assigned an interaction-based authority level governing how humans and AI collaborate:
| Level | Description | Example Activities |
|---|---|---|
| AI_AUTONOMOUS | AI executes without human involvement | Data aggregation, format validation |
| AI_PRIMARY | AI leads with human oversight | Intelligence preparation, trend analysis |
| HYBRID_AI_LED | AI generates, human reviews before action | COA development, risk assessment |
| HYBRID_HUMAN_LED | Human leads with AI assistance | Mission analysis, commander's guidance |
| HUMAN_ONLY | Human decides, AI may present information | Strike authorization, ethical judgments, risk acceptance |
Three activity categories are permanently locked to HUMAN_ONLY: AUTHORITY_DECISION, ETHICAL_LEGAL, and RISK_JUDGMENT. These locks are immutable in the smart contract.
Agent Integration
AI agents participate in DAO workflows under safety constraints:
- Agents draft and enrich proposals but cannot cast binding votes
- Agent recommendations are clearly labeled as AI-generated
- All agent outputs pass through human-in-the-loop checkpoints before affecting DAO state
- Agent DIDs provide verifiable attribution for every contribution