The policy is ahead of the tooling
In February 2020, the Department of Defense adopted five ethical principles for artificial intelligence: Responsible, Equitable, Traceable, Reliable, and Governable. The Responsible AI (RAI) Toolkit followed, requiring explainability for high-risk AI decisions and mandating that 'no AI solution will be operationalized by the DOD without explainability.' Traceability requires documenting all data and decisions including training data, processing methods, and outputs.
These are good principles. The problem is that the tooling to implement them at the tactical edge does not exist in any fielded system. Enterprise AI platforms can log model versions and query histories in cloud databases. Edge platforms running on disconnected laptops and phones have no equivalent infrastructure for model provenance, inference audit trails, or decision chain reconstruction.
The gap between policy and operational tooling is not hypothetical. It manifested in the Maven/Anthropic crisis of 2026, and it will manifest again every time a program office tries to field AI at the edge without a governance architecture.