Inspect the source. Challenge the answer. Keep verifiable evidence.
VeriTrooper examines source documents and structured data, tests AI answers against the material you approve, and preserves findings for human review. Run with local models, inspect deterministic checks, and verify signed evidence packages offline.
The exact interaction, claim, document condition, or uncertainty that failed the recorded test—not merely an aggregate score.
02 · Traceability
The tested input, observed output, verdict, governing source, scope, and limits needed to inspect the decision.
03 · Correction target
Evidence for deciding whether source material, retrieval, prompts, policy, model behavior, or human process needs attention.
04 · Retest the change
A repeatable run record that can be preserved beside the original after the system is corrected and tested again.
One audit system · three products
Audit the source. Audit the answers. Audit the system in operation.
Scout · before release
Build source-specific tests for calculations, precision, conditions, exceptions, and refusal behavior. Inspect the answers, stored references, and checks behind each result.
Q&A ledger with answers and source-linked references
Failure diagnostics and remediation priorities
Repeatable before-and-after audit record
Baseline, pipeline, and delta as scoped summary measures
Full paid outputs: system/data cards, applicable Annex IV record, crosswalk, and integrity material
Watchtower · after release
Audit samples of captured conversations or generated endpoint probes against your source material. Track recurring failures through scheduled reviews and retain the findings. Evaluation happens after delivery and does not block production actions.
Hourly through monthly or manual cadence
Monitoring Scope and Monitored Traffic Card
Flagged captured answers or probe results and supporting evidence
Full paid outputs: cycle history, diagnostics, machine exports, and seal-eligible records
SitRep · audit the source material
Compare claims across documents, check a document against supporting sources, and inspect structured records for inconsistent values, missing cells, and missing keys. See what was examined and what needs human review.
Unreadable, empty, duplicate, and damaged-file findings
Unsupported or contradicted claims
Cross-document conflicts, inconsistent values, missing cells, and missing keys
Suggested corrections where applicable, with recorded scope and limitations
Historical Scout walkthrough
See an audit move from source to finding.
This timed replay follows the actual Scout screens through a recorded 1,000-question GPT-5.5 audit against SEC 10-K filings. It is retained to show the product workflow; the current public evidence sample below is a separate 26 August 2026 gpt-5.6-sol run.
Governance is not a fourth product. It is the supporting workflow included with Scout, SitRep, and Watchtower. It combines each product’s verified evidence with system facts, classification decisions, oversight measures, and named human approvals.
Accountability remains human.The workflow supports review and documentation; it does not make legal determinations or approve deployment for the responsible organization.
Methodology
Inspect the decision. Investigate the cause. Retest the change.
Record the system, approved sources, configuration, roles, scope, and date.
Build or import source-grounded questions and reference answers.
Use deterministic checks for mechanically resolvable cases.
Record diagnostic opinions and optional verification separately from deterministic results.
Keep advisory disagreements and referrals visible alongside headline verdicts.
Trace each finding to its governing evidence and recorded conditions.
Use the evidence to investigate source, retrieval, prompt, policy, model, or process issues.
Preserve the item-level evidence, exclusions, review decisions, and any recorded signoff.
Re-audit after correction and preserve the new result beside the original.
Compliance evidence
Build the record before the regulator asks.
VeriTrooper supports EU AI Act technical-documentation, testing, accuracy, robustness, oversight, and monitoring workflows; NIST AI RMF Govern, Map, Measure, and Manage activities; and operational evidence for an ISO/IEC 42001 AI management system.
Annex IV testing, monitoring, and data records
System, data, scope, methodology, isolation, and framework documents
Measured results, limitations, exclusions, changes, and human decisions
Machine-readable records and tamper-evident retained packages
Evidence—not automatic certification.The customer and its advisers retain responsibility for system classification, legal interpretation, risk acceptance, conformity assessment, and regulator communications.
Security and deployment
Customer-controlled by design.
The locally deployed software does not automatically send customer audit data to VeriTrooper LLC.
Selected source files are read but never edited, moved, renamed, overwritten, or deleted.
Indexes, logs, captures, reports, and evidence packages are new local artifacts.
Local and fully offline operation is available and requires no provider key; the evidence records the boundary used for each run.
Remote model traffic goes to the endpoints selected by the operator. Trusted timestamp requests require a configured network connection.
Configured traffic goes directly to the customer-selected provider, not through VeriTrooper LLC.
Successfully completed, seal-eligible packages include signatures, manifests, and a portable offline checker.
Recommended local workstation: 32 GB RAM and 8 GB VRAM. Minimum memory is 16 GB. Windows evaluation installation requires at least 10 GB free, plus space for source indexes and audit records.
Current public evidence
Inspect a real audit record.
The public package exposes the report, item-level Q&A evidence, system and data context, framework mapping, and integrity material. Its measured result—88.08% with VeriTrooper versus 72.89% baseline across 889 scored questions—is one scoped outcome of the audit, not a general model rating or the product’s entire purpose.
Scout, SitRep, Watchtower, and Shared Governance Records for individual professionals, with local processing, evidence exports, and basic technical support.
Keep using purchased features without a subscription. Optional future major upgrades may be offered separately. Company deployment and guided assessments are quoted separately.
The Mac App Store edition is in preparation and pending Apple approval. App Store purchasing is not yet available.
The evaluation lasts 14 days from the first real audit, allows five total runs shared across Scout, SitRep, and Watchtower, and permits up to 200 questions or evaluated interactions per run.
Evaluation provides a limited, watermarked proof set for assessing the workflow. Full machine-readable records and extended evidence exports are included with the paid product; available artifacts vary by product and selected checks.
One complete ZIP for your platform.Request the free trial for Windows, Linux, Mac Apple Silicon, or Mac Intel. We will email the installer, checksum, and quick start instructions.
Veritrooper LLC is a Delaware limited liability company formed June 15, 2026. The technology is patent pending under U.S. non-provisional application 19/685,794. Guided pilots work directly with founder and sole inventor Brian L. Barbour.
Company, formation, ownership, assignment, patent-filing, and good-standing records are available during controlled diligence.
10-business-day guided assessment
Evaluate one important workflow.
One product, one use case, one approved source corpus, and up to 250 questions or monitored turns. Scope, pricing, and payment terms are established by written agreement before work begins.
Privacy and terms
Locally deployed VeriTrooper does not upload customer audit material to VeriTrooper LLC. Pilot submissions are processed through Formspree. VeriTrooper supplies testing and governance evidence, not legal advice or automatic certification. Public results remain limited to their recorded scope.