Asilla Selected for National Project 'GENIAC' by METI and NEDO

Asilla Inc. (Headquarters: Machida, Tokyo; CEO: Tsuyoshi Ono), with a vision of 'achieving a safe and comfortable world through the power of technology,' has been selected for the 'Post-5G Information and Communication System Infrastructure Enhancement R&D Project / Development of Competitive Generative AI Foundation Models (GENIAC),' organized by the Ministry of Economy, Trade and Industry (METI) and the New Energy and Industrial Technology Development Organization (NEDO).

Background
Since its founding in 2015, Asilla has focused on the research and development of 'Behavior Recognition AI' that understands human actions from video. In 2022, to address the severe labor shortage in the security industry, the company launched its AI security system, 'AI Security asilla.' As of May 2026, the proprietary security camera footage data accumulated through its operation has exceeded 8 million cases.

In April 2026, Asilla announced 'AsillaVision-v1,' an industry-specific proprietary VLM utilizing these data assets. Despite its lightweight 4B parameter design, AsillaVision-v1 achieves domain-specific performance in identifying abnormal behavior within facilities that surpasses major general-purpose large-scale VLMs.

Under this project, building on its proprietary data and R&D achievements, Asilla will develop security-domain-specific VLM foundation models in two tiers: a 4B model for edge environments and a 27B model for cloud environments. It will consistently promote social implementation in real environments. By enabling semantic understanding of abnormal events in the bird's-eye footage unique to security cameras, which cannot be captured by general-purpose models, Asilla aims to solve structural labor shortages in the security industry and strengthen the international competitiveness of Japan-developed AI foundation models.

Development Overview
Project Name: Post-5G Information and Communication System Infrastructure Enhancement R&D Project / Development of Competitive Generative AI Foundation Models (GENIAC)
Selected Theme: Research and Development of Security-Domain-Specific VLM Foundation Models Utilizing Large-Scale Proprietary Video Data

Outline of Development:
Utilizing its accumulation of over 8 million video data cases, Asilla will build security-domain-specific VLM foundation models of two different sizes tailored for specific purposes, and promote social implementation.

Development steps:
1. Dataset Construction: Build a training dataset from over 8 million cases of proprietary video data.
2. Foundation Model Development: Develop models in two sizes.
- 4B Model: For real-time processing in edge environments (local servers within facilities).
- 27B Model: For large-scale, high-precision analysis in cloud environments.
3. Demonstration and Implementation: Conduct Proof of Concept (PoC) in real-world commercial facilities to evaluate accuracy and implement socially.

About GENIAC
GENIAC (Generative AI Accelerator Challenge) is an initiative promoted by METI and NEDO aimed at strengthening Japan's generative AI development capabilities. It supports the development of foundation models, which are core technologies for various services, through the provision of computational resources and support for demonstration surveys for data and AI utilization.

Future Outlook
The foundation models developed in this project take a general-purpose approach to semantic understanding of security camera footage and are applicable regardless of region or language. Asilla plans to sequentially integrate the results of this project into its products, including 'AI Security asilla,' as a next-generation model for 'AsillaVision,' while advancing domestic deployment and collaborating with partner companies for overseas expansion.

Comments from Masahiro Wakasa, CTO of Asilla Inc.
'We are honored to have our proposal selected for this project. While the development of general-purpose models is progressing globally, building foundation models that are truly usable in industry-specific fields is a critical theme for Japanese AI development. This project aims to solve social issues by building highly practical foundation models in the security sector, where labor shortages are becoming critical, utilizing our proprietary data.'