Photographic truth, restored. Legal-grade authenticity for the AI era.

Where AI Provenance & Anti-Fake breaks today

Deepfake proliferation

Generative video and voice cloning have collapsed the cost of fabricating evidence to near zero — affecting elections, journalism, and the courts.

8× — increase in deepfake incidents 2023→2024 · Sumsub, 2024

Evidence rebuttal in court

Defendants increasingly challenge unsealed photographic and video evidence as 'possibly AI-generated', shifting burden onto prosecutors.

EU AI Act Art 50 — requires synthetic-content disclosure in regulated contexts

Synthetic identity at scale

AI-generated faces and voices enable account-creation and KYC fraud at industrial scale.

C2PA — is now adopted by Adobe, Sony, Leica, Microsoft, and the AP

Three patents, deployed against this industry's threat model

Each of Axowl's three filed patents maps to a specific structural failure mode in AI Provenance & Anti-Fake. Together they form a single, end-to-end defense.

Hierarchical Distributed Trust Fabric — Camera/phone chip (L1) · Platform (L2) · Registry (L3)

A PUF in the imaging chip seals the original frame at the moment of capture (L1). The hosting platform — newsroom CMS, evidence locker, social network — verifies and re-seals at L2. A regional provenance registry (L3) provides public lookup and conflict resolution. C2PA-compatible by design.

Transition-Sealed Integrity System — Every edit appended to a verifiable chain

Each edit (crop, color correction, AI enhancement, compression) is appended to the hash chain with the editor's identity. The viewer of the final image can audit the entire transformation history — was anything added? was a face swapped? was the timestamp shifted? — with cryptographic certainty.

Pre-grant LLM Conflict Verification — AI training and rights-of-use conflicts blocked at grant

A model-training permission such as "use customer-uploaded photos + produce synthetic likenesses" is recognized by the LLM gate as a conflict with consent terms and refused. Studios and platforms get policy enforcement that survives developer error.

Deployment that fits the threat model

Mobile deployments use the Defense tier integrated with phone SoCs. Newsroom and evidence-locker deployments use AWS F2 sealing servers. Long-term volume integrates a dedicated provenance ASIC in the imaging pipeline.

Recommended tier: T2 · Defense (mobile chip) → ASIC at scale

Deployment path: F2 sealing server · Mobile SoC integration · C2PA registry federation

Operational detail: Newsroom and platform sealing runs on AWS F2 within 72 hours. Mobile integration begins with reference-design SDK on Snapdragon / Tensor / Apple silicon, with a long-term path to dedicated provenance ASIC. C2PA registry federation supports cross-platform provenance lookup.

Three concrete deployments

News photography provenance

A wire photographer's camera seals every frame at capture. The newsroom verifies the seal; the published image carries an embedded provenance chain that any reader can audit.

Court evidence chain

Body-cam and bystander video sealed at capture survive defense challenges based on AI-fabrication claims; the chain proves what the lens actually saw.

Platform authenticity badges

Social platforms display an authenticity badge for content with a verifiable C2PA-compatible chain, restoring viewer trust without imposing on creators.

Versus what's deployed today

Today — EXIF + watermark

EXIF is freely editable. Watermarks are routinely defeated by re-encoding. Neither survives an adversarial editor.

With DPSM — Axowl DPSM

PUF-anchored seals cannot be regenerated by any party that does not physically hold the original capture device. The chain is verifiable without any trust in the publisher.

Standards & regulatory frameworks aligned

• C2PA 2.0
• EU AI Act Article 50
• NIST AI Risk Management Framework
• Adobe CAI
• ISO 22144 (planned)
• IPTC Photo Metadata