An open, worldwide standard for real-time deepfake detection.
Vendor-neutral. Volunteer-led. Catalyzed by industry, owned by no one.
The Problem
Deepfakes have crossed from curiosity to systemic threat to human trust. Fraud, impersonation, election interference, image-based abuse.
The Standard
Seven first-principle requirements every conforming detection product must meet. Real-time, on-device, hardware-agnostic, privacy-preserving.
The Mark
Once banks, governments, and platforms require this mark in procurement, every vendor must certify or be excluded.
Founding Signatories
Coming soon. Sign the charter to join the founding cohort.
Seven Founding Principles
Real-time
Detection at the point of interaction, not post-event forensic upload. Synthesis happens live; detection must too.
On-device capable
Runs locally without mandatory cloud dependency. Sovereignty and latency demand this.
Air-gap compatible
Functions in offline or isolated environments. Critical for defense, healthcare, finance.
Privacy-preserving
Analysis can occur without media leaving the device. Compliant with GDPR, HIPAA, and regional data sovereignty law.
Hardware-agnostic
No lock-in to a single chip or vendor ecosystem. Portable across architectures.
Transparent scoring
Authenticity confidence is reported, not hidden. Response actions remain configurable per the deployer's governance policy.
Independently testable
All claims must be verifiable by a neutral benchmark. No self-reported accuracy.
How the standard becomes infrastructure
- 01
Threat Intelligence
Neutral pool of synthesis techniques, attack patterns, and adversarial samples.
- 02
Independent Testing
Academic lab benchmarks detectors against the intel pool.
- 03
Certification
Conforming products earn the ARIA Certified mark.
- 04
Regulation & Insurance
Detection written into financial, KYC, procurement, and insurance frameworks.
- 05
Default Adoption
Detection becomes infrastructure. The loop closes.