AUTHCHECK
The problem
How it works
Governance
Founders
Request a pilot
SOVEREIGN TRUST INFRASTRUCTURE · ABU DHABI
AI can now generate video no human eye can question. AuthCheck verifies what's real, what's generated, and what's being passed off as something it isn't — for newsrooms, platforms, and courts.
Request a pilot
See how it works
CHAIN OF CUSTODY
SHA-256 ····
SCANNING
SORA 2 · OCT 2025
VEO 3.1 · 4K NATIVE
KLING 3.0 · MULTI-SHOT
RUNWAY GEN-4.5 · REAL-TIME
SEEDANCE 2.0 · 2026
PIKA · SOCIAL-SCALE
HAILUO · MINIMAX
NEW GENERATORS LAUNCH MONTHLY —
THE PROBLEM
Every era of generated video got cheaper, faster, and harder to catch. The tools below aren't research demos — they're consumer apps.
GEN CLIP · 4 SEC
2022
Four seconds of blurry motion. Warped hands, swimming backgrounds. Anyone could spot it.
DETECTABLE BY EYE
GEN CLIP · 60 SEC
2024
Photoreal minute-long scenes with coherent physics. The tells got subtle: lighting logic, temporal flicker.
DETECTABLE BY EXPERTS
GEN CLIP · 4K + AUDIO
2026
Veo 3.1, Kling 3.0, Sora 2: 4K output, native audio, multi-shot consistency. The human eye is out of the game.
MACHINE-ONLY DETECTION
The scale is not hypothetical. The Sora app passed a million downloads in under five days at launch. Generation pricing keeps collapsing toward a rounding error. Volume compounds monthly — and every institution that publishes video inherits the problem.
HOW IT WORKS
Every video is examined by independent layers, each honest about how much it can prove. Cryptographic proof outranks everything. Estimation is labeled as estimation.
CERTAINTY
C2PA content credentials verified by pure math. A valid signature proves a file untouched since capture; a broken one proves tampering. Deterministic — the layer a court can stand behind.
ESTIMATION
A continuously retrained detector ensemble reads the statistical fingerprints generators leave in pixels and motion — and reports a calibrated risk score with confidence intervals, never a verdict.
CORROBORATION
A rule engine over the file's hidden records: missing sensor data, impossible timestamps, re-encode laundering, generator signatures. Cheap signals that catch careless fakes constantly.
The biggest lie online isn't a generated video — it's a real one with a false story. A 2015 fire reposted as this week's breaking news. AuthCheck cross-references footage against its earliest appearance on the web and against what reputable sources actually reported — catching reuse and misattribution no pixel-detector can see.
Every verification ends in evidence: a SHA-256 chain-of-custody fingerprint, per-layer findings, and a plain-language summary an editor reads in ten seconds.
THE MOAT
WATCH
new generators
COLLECT
fresh samples
RETRAIN
fine-tune models
BENCHMARK
human-gated
DEPLOY
days, not months
✓
Anyone can download a base model.
Nobody can download a head start of fresh training data accumulated across every generator since day one.
✓
Days to respond, not months.
When a new generator launches, our pipeline collects its output, retrains, benchmarks, and ships an update while public detectors stay frozen in time.
✓
The dataset compounds.
Every verification and every generator release deepens an asset that cannot be replicated retroactively. The moat grows whether or not any single deal closes.
Continuous retraining is why the defense stays current — and why customers subscribe instead of buying once.
WHY NOW
Labeling and provenance rules are landing on every publisher of video. The infrastructure to comply doesn't exist yet — that's the gap AuthCheck fills.
IN FORCE · AUG 2026
AI-generated content must be labeled. Platforms and broadcasters operating in Europe need verification infrastructure to comply.
IN FORCE · AUG 2026
Provenance disclosure requirements for AI content reach the largest media market in the United States on the same timeline.
ACCELERATING · GCC
The UAE and the wider region are building national AI governance. Media verification is a sovereign capability — not one to license from abroad.
Regulators will mandate that video be verifiable. They will not mandate whose system verifies it. AuthCheck is built to be that system — owned in the region, aligned with its strategy.
WHO IT SERVES
ENTRY WEDGE
Verify user-generated and agency footage before air. One published fake costs a reputation; a sixty-second check costs nothing.
COMPLIANCE SCALE
Label AI content at scale as disclosure laws land. API-first verification built for volume and audit trails.
EVIDENCE GRADE
Forensic reports with chain of custody, cryptographic findings first, estimation clearly labeled — built to survive cross-examination.
CO-FOUNDERS
OD
CO-FOUNDER
Robotics engineering graduate with R&D experience at Bristol Robotics Laboratory. Built AuthCheck's verification pipeline, training loop, and evidence reporting end to end — before a single dollar was raised.
HK
CO-FOUNDER
Student at Sorbonne Abu Dhabi, majoring in AI and Mathematics for Data Science. Driving AuthCheck's partnerships, dataset operations, and pilot program — turning a working prototype into the region's verification layer.
REQUEST A PILOT
We're onboarding a small number of newsrooms, platforms, and institutions as founding pilot partners. Bring us your hardest footage.
AUTHCHECK · ABU DHABI, UAE · PILOTS OPENING 2026
Evidence, graded by certainty.
© 2026 AUTHCHECK · ABU DHABI, UAE