Turn representation-layer analysis into an operational program covering assets, trust boundaries, controls, maturity, evidence, metrics, governance, tabletop exercises, and incident response.
Machine Tradecraft defense secures the boundary between untrusted artifacts and privileged AI systems by preserving raw evidence, comparing representations, constructing provenance-labelled canonical input, isolating capabilities, requiring human confirmation, and measuring control performance across separate maturity dimensions.
The submitted report remains byte-identical. This separate review, checked 2026-08-26, narrows implementation claims and gives visitors a decision path before they generalize from a local result.
Output produced deterministically by the local runtime.
Useful model or subset that must not be mistaken for full conformance.
Conclusion that needs an exact implementation, trust system, model, parser, or human review.
| Question | What the local page can answer | What it does not establish | Next evidence step |
|---|---|---|---|
| Which defensive dimension needs the next evidence-backed action? | The planner identifies one local priority from the selected inputs. | Actual control efficacy or organizational risk. | Collect evidence and test the dimension independently. |
| What is the program's maturity level? | Nine separate illustrative rows are shown. | A certified maturity level or comparable benchmark. | Use an authorized assessment method and documented criteria. |
| Will the roadmap reduce risk? | The page supplies planning orientation only. | Outcome, cost, timing, or residual risk. | Define measures, owners, baselines, and review intervals. |
Shows how evidence gaps and next actions differ without producing one opaque score.
Expected boundary: The planner returns nine separate dimensions, a clear non-certification notice, and no persistence.
Open prepared laboratoryNo single view is automatically authoritative. Preserve the source, identify each parser or receiver, and compare their outputs before authorizing a consequential decision.
Content, parsers, models, tools, credentials, stores, logs, and provenance records.
Representation transitions and the authority reachable from each channel.
Acquisition, validation, comparison, canonicalization, isolation, confirmation, and egress.
Architectures, configurations, tests, run records, exceptions, and incidents.
Separate observable dimensions rather than one combined score.
Classification, preservation, containment, remediation, recovery, and lessons learned.
The workflow preserves evidence before transformation and keeps structural inspection separate from execution, remote verification, or model behavior.
These controls are contextual. They reduce a defined risk; they do not guarantee safety, truth, attribution, or resistance to every adaptive attack.
Keep the reader network-denied, uncredentialed, and unable to change state.
Construct a new canonical model input rather than mutating untrusted objects in place.
Measure extraction, uptake, false positives, transformation resilience, and fail-closed behavior separately.
Require human confirmation for consequential or externally visible actions.
Treat framework mappings as references, not certification or endorsement.
Definitions are linked into the site-wide glossary and back to the full report.
A threat model centered on transformations among bytes, structures, rendered views, tokens, and actions.
The observable degree to which a control is defined, implemented, measured, and improved.
A secondary measure reducing risk when the preferred primary control is unavailable.
A concrete artifact required to support a control or maturity claim.
Risk remaining after selected controls are applied.
The measured portion of the defined attack surface addressed by tested detection logic.
A control’s measured ability to retain performance after specified input transformations.
The bounded evidence set recording source hashes, parser versions, transforms, decisions, and outcomes.
A network-denied, uncredentialed component that extracts facts from untrusted artifacts without execution authority.
These links are provided for visitors who want the governing specification, paper, framework, or implementation documentation. Links open in a new tab; the site does not fetch them during runtime analysis.
Vulnerability severity scoring, explicitly distinct from program maturity.
Normative or first-party specification material.Govern, identify, protect, detect, respond, and recover context.
First-party guidance, framework, registry, or standards-program material.AI governance, mapping, measurement, and management.
First-party guidance, framework, registry, or standards-program material.Practical actions supporting AI RMF outcomes.
First-party guidance, framework, registry, or standards-program material.Adversarial AI tactics, techniques, and mitigations.
First-party guidance, framework, registry, or standards-program material.Software-assurance maturity practices.
First-party guidance, framework, registry, or standards-program material.Taxonomy for AI attacks and mitigations.
First-party guidance, framework, registry, or standards-program material.Secure defaults and manufacturer responsibility.
First-party guidance, framework, registry, or standards-program material.The full submitted report is preserved byte-for-byte in the governed research library and in durable repository documentation. The laboratory turns selected concepts into deterministic local output without external calls or hidden persistence.
An operational architecture covering assets, representation-layer threats, reader/executor boundaries, control families, maturity dimensions, evidence, metrics, governance, tabletop exercises, incident handling, roadmaps, and local assessment.
Read governed reportRecord explicit maturity selections for nine independent dimensions, preserve one bounded evidence note, and generate a transparent local action plan without an overall score or certification claim.
Open bounded laboratory