Projected U.S. AI-enabled fraud by 2027 — a market forming right now.
Exhibit 01 — Kim Boher · Current Venture
DeFake
Founded January 2026 · Delaware C-Corp · San Francisco.
Multimodal forensic AI fraud detection — across voice, video, and images — for insurance, legal, media, and trust & safety teams. Digital evidence is becoming unreliable faster than institutions can adapt.
Over the next decade, every insurer, marketplace, financial institution, law firm, and government agency will need infrastructure to verify whether media is authentic, manipulated, AI-generated, or misrepresented — and produce defensible evidence, not merely a detection score.
Founder & CEOKim BoherCompanyDeFake · Delaware C-Corpdefakes.com · U.S. patent pending
The agenda
Why now
We rely on photos, videos, and documents to make critical decisions — and that foundation of trust is breaking. Tools like Sora, Kling, and Runway let anyone generate convincing synthetic content, and AI agents now create and distribute it at scale. When anything can be faked, proof of what is real becomes critical infrastructure. That is the agenda behind DeFake — proof of reality in the AI era.
Of today's AI-generated video that legacy, face-swap-era detectors miss.
What existing tools return: a probability — no explanation, nothing to verify.
The convergence
Why me
Over a decade building fraud detection inside government, the Israel Securities Authority, and major financial institutions, I hit sophisticated fraud — $10M in transactions — that existing tools were never designed to catch. And a detection score isn't enough for compliance: you have to document the methodology, explain the analysis, and show why the evidence supports the conclusion. That gap is why I built DeFake.
Fraud & ComplianceBuilt the detection algorithms inside a securities regulator and major financial institutions
+ML & Computer VisionFull-stack, deep learning and blockchain — shipped to production
=DeFake
Fraud domain — 15+ years
- Israel Securities AuthorityMarket-manipulation detection algorithmsIsrael's securities regulator — the SEC equivalent
- Harel Insurance & FinanceFraud, compliance, risk and real-asset valuationReal-estate investment control
- Ayalon FinanceInvestment control and fraud monitoringMutual-fund oversight
- Ministry of JusticeFinancial oversight and government caseworkRegulatory frameworks
- Hebrew UniversityMulti-layer security operationsWhere adversarial thinking started
Builder track
- ML & computer visionDeep learning, NeRFs, neural-artifact detection, LLM agents, voice pipelines, TinyML.
- Full-stack developmentFrontend: TypeScript · React · Next.js · Angular · Tailwind. Backend: Node.js · Python · FastAPI · REST & GraphQL APIs. Data: PostgreSQL · Supabase · MongoDB. Cloud & deployment: AWS · GCP · Vercel · Railway · Docker · CI/CD. Systems: Rust.
- Blockchain & provenanceRust smart contracts · NEAR · Base L2 · C2PA · cryptographic anchoring.
The product
Suspect media in,
proof delivered
Existing tools were built for an earlier generation of deepfakes — and even when they flag something, they return only a probability score. DeFake is different: instead of a single black-box model, six independent detection and investigation methods each examine a different layer of the content — producing not just a verdict, but a traceable, evidence-based forensic report.
Suspect media in
Fraudulent insurance claim
Stolen identity video
Fake news footage
Deepfaked court evidence
Synthetic KYC identity
DeFakeAnalysis
Media integrity
Origin & provenance
Tamper analysis
Proof, delivered
Forensic report
Court evidence
Financial damage evaluation
Takedown notice
Verdict
Backing & credentials
Accelerators,
patent & benchmark
Harvard Foundry Deep Tech Incubator
Completed · 2026
Founders Inc — Canopy
Accepted · 2026
Google for Startups Cloud Program
Approved · 2026
Markets
One engine,
many high-stakes markets
The same question — is this media authentic, manipulated, AI-generated, or misrepresented? — is worth billions across every institution that runs on digital evidence.
$308Bannual U.S. insurance fraud
Insurance
Claims photos, manipulated evidence, and court-ready forensic reports.
$300B+annual U.S. fraud (estimate)
Health care
Evidence authentication, claims documentation, and abuse detection.
$108BAI fraud-detection market by 2033
Fintech & crypto
AI selfie, voice, and identity-media checks inside KYC / KYB.
$35Bfraud-prevention market, 2025
Marketplaces
Fake listings, staged damage, product-image fraud, and payout disputes.
$9Bdeepfake-tech market, 2025
Creator & media platforms
Impersonation, non-consensual synthetic content, and identity protection.
Court-gradechain-of-custody evidence
Legal & investigations
Verify photos, video, and audio before they reach a filing or a courtroom.
Build proof
Every layer of DeFake,
built somewhere first
The forensic engine is not a first attempt at any of its parts. Computer vision, on-chain provenance, agent architecture, realtime pipelines and edge inference each shipped in an earlier production system.
AR Digital Twin
Phone video to production 3D — the computer-vision pipeline behind the forensic frame analysis.
Open case file →Voice-to-Voice AI Assistant
A realtime voice pipeline built before realtime voice APIs existed.
Open case file →Cryptographic Chain of Custody
On-chain evidence anchoring — the provenance layer DeFake runs on.
Open case file →TinyML on Microcontrollers
Full ML lifecycle inside 256 KB of RAM — dataset to quantized on-device inference.
Open case file →Verity
An AI attester scoring quality, activating a smart contract, moving payment automatically.
Open case file →Cognitive Intelligence Labs
LLM agents for legal evidence workflows.
Open case file →