Program managers & mission owners
Define the evidence required before fielding, set acceptance thresholds and operating limits, and document residual-risk decisions.
Defense & government
AI evaluation and assurance for mission systems.
We test AI-enabled systems under mission-relevant and adversarial conditions. We document how the system behaved and give program teams the material they need to set controls, acceptance criteria, escalation paths, and review conditions.
How the work fits
Syntony's defense and government practice connects technical evaluation with the officials responsible for procurement, authorization, fielding, risk acceptance, and ongoing operation.
Evidence for procurement, authorization, and fielding
Public institutions already have principles and review bodies. They still need evidence about how a particular AI-enabled system behaves in its intended environment. Program teams use that evidence to decide when the system can operate and under what conditions.
Federal AI assurance draws on AI safety, cybersecurity, acquisition, and test and evaluation. Red-team and other test results become useful when they inform acceptance criteria and lifecycle risk decisions.
Syntony connects the technical results to those program decisions.
Built for
Teams that make or support the decision
The work is designed for teams responsible for technical performance, security, procurement, oversight, or mission outcomes.
Acquisition & contracting teams
Turn vendor claims into testable requirements, evaluation plans, performance measures, and contract obligations that remain enforceable after award.
AI, data, cyber & governance offices
Manage AI portfolios through controls and review gates, then maintain monitoring, incident-response, and change-management processes.
Test & evaluation organizations
Evaluate the base model, the integrated system, the human-system team, and operational performance.
Defense AI vendors & integrators
Prepare evidence for buyer evaluation, cybersecurity review, authorization, and continued delivery.
Primes & allied public-sector teams
Assess AI-enabled suppliers and align evidence across organizations, missions, review processes, and operating environments.
Services by program stage
Start with the decision the program needs to make
Before the pilot
Strategic Risk & Readiness
Define the mission, system boundary, threat model, governance gaps, and evidence needed before a pilot begins.
View service → Before fieldingMission AI Assurance & TEVV
Apply test, evaluation, verification, and validation to the model, its integration, the human-system team, and operational behavior.
View service → For agentic workflowsAgent Security & Control
Test prompt injection, permissions, tool use, identity controls, delegation, override, and recovery in agentic workflows.
View service → During acquisitionAcquisition & Vendor Evaluation
Translate requirements into evaluation protocols, compare vendors under realistic conditions, and define acceptance and monitoring criteria.
View service → For review and authorizationGovernance & Authorization Readiness
Organize threat, provenance, test, control, monitoring, and risk-acceptance evidence for program review and authorization.
View service → After deploymentContinuous AI Assurance
Keep tests, evidence, accepted risks, controls, and change reviews current throughout the system lifecycle.
View service →
Typical deliverables
Engagement outputs
Findings are tied to the program artifacts, controls, acceptance criteria, escalation rules, and regression tests needed to act on them.
- Mission-specific risk and threat model
- Scenario and adversarial evaluation plan
- Replayable traces and test results
- Model, system, and human-workflow findings
- Control and ownership map
- Acceptance and escalation criteria
- Residual-risk and program decision brief
- Regression suite for future changes
How an engagement runs
From mission definition to re-testing
- 01Define the mission
Identify the system, operator, operating environment, relevant adversary, and decision authority.
- 02Set the evidence requirements
Choose evaluation layers, scenarios, data, thresholds, and acceptance criteria.
- 03Challenge the system
Test behavior under representative, adversarial, and degraded conditions.
- 04Support the decision
Document findings as controls, residual-risk statements, and review-ready artifacts.
- 05Re-test material changes
Run the tests again when the model, data, integration, mission, or threat materially changes.
Framework alignment
Map evidence to the required review process
When useful, Syntony maps findings and evidence to NIST AI RMF, the DoD AI Cybersecurity RMF Tailoring Guide, applicable cybersecurity and risk-management processes, responsible-AI guidance, acquisition artifacts, and customer-defined test and evaluation requirements.
The mapping supports a specific review. It does not establish universal compliance or certification.
Relevant foundation
Relevant experience
Fourteen years across geopolitics and emerging technology.
Prior decision-science work supporting U.S. Department of Defense clients on emerging-technology risk, AI integration, and security cooperation.
Frontier-model red-team experience paired with research on international humanitarian law, synthetic media, European defense cooperation, and governance lag.
Founder-led delivery with specialist collaborators engaged where appropriate and with client agreement.
Discuss a program
Tell us which decision the evidence needs to support.
We can scope work around a vendor evaluation, pilot, agentic workflow, authorization package, or assurance program.