Cryptographic provenance can establish facts about origin/history and integrity of assertions; it does not prove the depicted event or statement is factually true.
MODLAN V91 · DIGITAL PROVENANCE + SYNTHETIC MEDIA ASSURANCE
Trust the evidence. Interrogate the artifact.
V91 establishes a forensic-grade artifact intelligence layer for images, documents, PDFs, office files, email, audio, video, archives, text/code and unknown binaries. It combines cryptographic identity, provenance, deterministic file forensics, watermark/provider signals, multimodal synthetic-media detection, cross-artifact correlation and governed human review—without reducing authenticity to one opaque AI score.
FORENSIC DOCTRINE
Precise conclusions survive challenge.
V91 is designed to support investigators and analysts with explainable, reproducible evidence while preventing a probabilistic detector from silently becoming a fraud, authenticity, identity, or legal conclusion.
Absence of a supported watermark or Content Credential does not establish that content was created by a human or never edited by AI.
Synthetic-media detectors are versioned probabilistic evidence sources. Material findings preserve per-engine results, uncertainty, calibration and counter-signals.
Legitimate cropping, transcoding, enhancement, redaction and document processing can alter forensic features. Context and authority remain necessary.
A hash proves byte identity against the hashed artifact; it does not establish who created the content or whether the content depicts reality.
Unsupported or missing analysis remains indeterminate. MODLAN does not convert missing detector coverage into a clean result.
A material insurance, investigative, legal or fraud conclusion must not rely solely on an opaque automated synthetic-content score.
MODLAN may consume approved GHOST™/FEOS™ findings while CST scoring logic, TTP mappings, model internals and unpublished embodiments remain protected.
ENGINEERING BASIS
Open standards outside. Protected intelligence inside.
NIST AI 100-4
Layer provenance/authentication, watermarking/labeling, detection, testing, auditing and maintenance rather than relying on one detector.
NIST AI RMF + GAI Profile
Govern/map/measure/manage detector risks, limitations, human oversight, evaluation and documentation.
C2PA Content Credentials
Validate signed provenance assertions, ingredients, edits and AI-related source declarations without treating provenance as truth.
Insurance AI governance
Keep AI-assisted insurance workflows governed, documented, reviewable and compliant with applicable insurance requirements.
Secure file handling
Quarantine untrusted artifacts, identify content from bytes, isolate parsers, prevent execution and bound archive/resource expansion.
DIGITAL PROVENANCE COMMAND
One artifact. Independent evidence planes.
Explore the V91 architecture using synthetic scenarios only. This public surface performs no uploads, biometric matching, malware execution, production evidence queries, or automated insurance decisions.
JPEG · PNG · WebP · HEIC · TIFF
DETECTORS- C2PA/credential
- watermark/provider
- synthetic-image ensemble
- edit/manipulation
- EXIF/XMP
- magic/container
- quantization/recompression
- perceptual fingerprint
- source/device consistency
Hash the received artifact, bind intake context, preserve immutable identity, and distinguish original from derived copies.
SCOPE all bytes/streams
OUTPUT SHA-256 + intake event + derivative lineage
BOUNDARY · Integrity proves byte identity, not truth of depicted content.AUTHORITY
- authorized matter/tenant
- lawful purpose
- operator/service identity
JPEG loss photo
Image contains generation-tool metadata and an AI-provider provenance assertion.
- credential/provider signal
- metadata generation hint
- image ensemble result
- cross-artifact reuse
RESPONSE Preserve original, validate provenance independently, run image forensic ensemble, correlate authorized claim context, and route material findings to analyst review.
DO NOT CONCLUDE · Do not infer fraud solely because AI generation/editing is detected.MULTIMODAL ASSURANCE MATRIX
Images are only one evidence class.
V91 applies a common intake/custody architecture while dispatching content to modality-appropriate forensic and synthetic-media analysis. Unsupported formats remain indeterminate rather than silently clean.
Image Forensics
JPEG · PNG · WebP · HEIC · TIFF
ARTIFACT LINEAGE GRAPH
Detection is evidence in a chain—not a magic answer.
SOURCEAcquisitionauthorized source + original bytes
Integrity, provenance, synthetic-media indicators, manipulation indicators and factual truth remain separate questions. V91 preserves those distinctions all the way to reporting.
OUR OWN DETECTION ARCHITECTURE
MODLAN owns the orchestration, evidence model and deterministic forensic substrate.
The V91 server substrate now performs bounded byte-level intake, SHA-256 fingerprinting, magic-byte classification, extension/MIME mismatch detection, entropy measurement, active-content indicators, archive safety gating, provenance-marker discovery and AI-tool metadata hints without executing the submitted content. Modality-specific model adapters can add calibrated image, voice, video, text and document detectors as they are validated.
CST GHOST™ / FEOS™ BRIDGE
Reuse intelligence. Do not duplicate the invention.
MODLAN may consume approved findings from the existing CST GHOST™ synthetic-media and FEOS™ forensic-evidence systems through a protected interface. CST scoring logic, INSITE/TTP mappings, model internals, protected algorithms and unpublished embodiments remain confidential and are not recreated in the public MODLAN codebase.
INSURANCE DECISION BOUNDARY
A detector is evidence—not an adverse-action engine.
V91 is designed to surface forensic indicators, provenance and uncertainty for authorized review. A synthetic-media signal does not, by itself, establish fraud, misrepresentation, identity, coverage, liability, admissibility, causation, damages, or any other legal/insurance conclusion. Material decisions remain governed by applicable authority, evidence, policy, law, human review and documented reasoning.
NEXT BUILD · V92
Business Continuity.
V92 exercises continuity across the increasingly critical MODLAN operating plane, including artifact-intelligence dependencies, evidence preservation, recovery communications, alternate procedures, catastrophe surge and carrier/partner integrations. Performance and capacity requirements remain part of continuity, staging and launch qualification.