#Japan #AI Data #Minato City #NII #SYNTHETIQ VISION

Addressing the Threat of Video Fraud

In recent years, the evolution of generative AI has made it possible for anyone to create fake images and audio that are virtually indistinguishable from the real thing. This technology is being misused in various ways—bypassing online identity verification (eKYC), spreading misinformation on social media, and producing deepfake pornography using children’s images, leading to severe real-world consequences.

The once straightforward assumption that video evidence is objective has started to crumble. As a result, determining how to verify the authenticity of video content, which is increasingly challenging to discern by mere observation, has become an urgent task for society, government, and businesses alike.

Establishing a Fake Detection Foundation

AI Data, a company based in Minato City, Tokyo, has partnered with the National Institute of Informatics (NII) to address these pressing issues. They are using NII’s fake video detection AI, SYNTHETIQ VISION, as foundational technology to construct an in-house 'Fake Video Detection Platform' tailored for potential use cases such as eKYC, social media, and child protection. Validation of the platform’s effectiveness and operational methods has now commenced.

As the validation process continues, AI Data will evaluate specific forms of service delivery and targeted areas for the future.

What is SYNTHETIQ VISION?

SYNTHETIQ VISION is an advanced fake video detection AI developed by the National Institute of Informatics. Some of its key features include:

• - The AI analyzes unique “subtle generation traces” characteristic of forged or synthesized facial videos, outputting a score indicating the likelihood of authenticity.
• - It employs a numeric evaluation approach for videos that are challenging for the human eye to distinguish.
• - The model is designed to maintain detection performance even after video compression or processing.
• - It features an architecture capable of preserving judgment results and associated logs, facilitating later verification and auditing processes.

AI Data aims to assess:

• - The potential of SYNTHETIQ VISION in actual business scenarios, including eKYC, responses to social media reports, and child protection.
• - The most effective ways to integrate it with existing visual checks and fraud detection systems.
• - Strategies for managing and preserving judgment logs in light of evidentiary and accountability concerns.

It is important to note that this press release does not guarantee the performance of SYNTHETIQ VISION, but announces the initiation of validation using this technology.

Anticipated Use Cases and Integration Targets

The fake detection platform is envisioned for use in several areas, contingent on the validation progress and discussions with relevant stakeholders. Possible applications include:

• - eKYC / Financial Security:

- Aiding in the detection of identity fraud during account openings and fraudulent transactions using recorded videos or AI-generated faces.
- Evaluating the potential integration of a two-step identity verification process with existing facial recognition AI.

• - Social Media Content Review / Video Streaming Platforms:

- Automating pre-screening processes for user-uploaded videos and live broadcasts.
- Exploring ways to provide technical evidence for deletion requests or content suspension based on reports received.

• - Child Protection / Educational Settings:

- Assisting in the detection of deepfake pornography or bullying videos involving children.
- Evaluating the integration into initial response processes in collaboration with schools, educational boards, and law enforcement.

These scenarios represent initial considerations, and specifics regarding implementation and operation will be determined through ongoing validation outcomes in cooperation with institutions across each field.

Prospective Offering Models

Looking ahead, various models for external offerings are under consideration, but these remain speculative during the validation phase:

• - SaaS Model:

- Users uploading videos through a web interface to receive scores and logs indicating potential fakes, reevaluating usability, processing performance, and security requirements.

• - API Integration Model:

- Developing API interfaces for connecting to existing systems in financial institutions, municipalities, and businesses, while assessing technical specifications, throughput, authentication methods, and log management.

• - OEM / Resale Model:

- Investigating options for integrating the fake video countermeasure solution under various brand umbrellas, considering business schemes, support structures, and delineation of responsibilities.

Future Outlook

AI Data aims to implement a technological defense against the proliferation of deepfakes through effective use of SYNTHETIQ VISION and expanding use cases. Future initiatives will include:

• - Planning and executing validations and proofs of concept related to eKYC, social media, education, and municipal operations.
• - Contributing to the establishment of operational models, guidelines, and rules in partnership with public and private sectors and research institutions.
• - Exploring the integration of this initiative with its data recovery and digital forensic services for a cohesive investigation and evidence preservation framework.

AI Data aspires to establish itself as a vital part of the AI safety net supporting the detection and prevention of fake videos.

About AI Data Inc.

• - Name: AI Data Inc.
• - Founded: April 2015
• - Capital: 100 million JPY (with 1.525 billion JPY in capital reserves)
• - CEO: Takahito Sasaki
• - Location: 4th Floor, Metro City Kamiyacho Building, 5-1-5 Toranomon, Minato City, Tokyo
• - Website: AI Data Official Site

For over 20 years, AI Data has focused on protecting and utilizing data assets for enterprises and individuals through its data and intellectual property infrastructure. Trusted by over 10,000 companies and more than one million customers, AI Data has received the top sales award in the data ecosystem sector for 16 consecutive years. Its data infrastructure provides cloud data management and recovery services whilst earning recognition for forensic investigations and evidence disclosure in legal sectors.

In the intellectual property sector, it supports IP management and monetization through its group company’s systems and marketplace. By integrating these functions with the AI-powered platform AI Kongming™, AI Data is committed to enhancing social infrastructure through diligent data management and IP protection.