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

#
News

Asilla, Inc. (Headquarters: Machida, Tokyo; Representative Director & CEO: Tsuyoshi Onoue; hereinafter "Asilla"), whose vision is "a safe and comfortable world through the power of technology," announces that it has been selected for the "Research and Development Project for Strengthening Post-5G Information and Communication Systems Infrastructure / Development of Competitive Generative AI Foundation Models (GENIAC)," an initiative led 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 research and development of "action recognition AI" that understands human behavior from video footage. In 2022, the company launched "AI Security asilla," an AI-powered security system leveraging this technology to address the serious labor shortages facing the security industry. The proprietary security camera footage data accumulated through its operations (※) has exceeded 8 million records as of May 2026.

In April 2026, Asilla announced "AsillaVision-v1," an industry-specific proprietary VLM built on this data asset. Despite its lightweight design of only 4B parameters, AsillaVision-v1 achieves domain-specific performance surpassing major large-scale general-purpose VLMs in identifying anomalous behavior within facilities.

In this project, building on its proprietary data and R&D track record, Asilla will develop a security domain-specific VLM foundation model in a two-model configuration — a 4B model for edge environments and a 27B model for cloud environments — and will drive the initiative through to real-world social implementation. By enabling semantic understanding of anomalous events in the overhead footage characteristic of security cameras, which general-purpose foundation models struggle to capture, Asilla aims to address the structural challenge of severe labor shortages facing the security industry and contribute to strengthening the international competitiveness of Japan-originated AI foundation models.

※ All data is collected and used with the consent of the deploying facilities and has been anonymized.

Project Overview

Program Name:

Research and Development Project for Strengthening Post-5G Information and Communication Systems Infrastructure / Development of Competitive Generative AI Foundation Models (GENIAC)

Selected Theme:

Research and Development of a Security Domain-Specific VLM Foundation Model Utilizing Large-Scale Proprietary Video Data

Development Overview:

This project will leverage Asilla's proprietary dataset of over 8 million records to develop security domain-specific VLM foundation models at two scales suited to different use cases, with a commitment to driving the work through to real-world social implementation.

Development will proceed in the following steps:

Dataset Construction: Building a training dataset from over 8 million proprietary video records

Foundation Model Development: Developing security domain-specific VLM foundation models at two scales for different use cases

  • 4B Model: For real-time processing in edge environments (on-premises facility servers)
  • 27B Model: For large-scale, high-precision analysis in cloud environments

Validation & Social Implementation: Conducting real-environment accuracy evaluation and proof-of-concept (PoC) for social implementation at commercial facilities operated by validation partner companies

About GENIAC

GENIAC (Generative AI Accelerator Challenge) is a program promoted by METI and NEDO with the goal of strengthening Japan's domestic generative AI development capabilities. The project aims to sustainably enhance the domestic capacity to develop generative AI foundation models — the core technology underpinning a wide range of AI-powered services — through support such as providing computational resources for foundation model development and facilitating practical research on data and AI utilization.

URL: https://www.meti.go.jp/policy/mono_info_service/geniac/index.html

Future Outlook

The foundation models developed through this project take a general approach to semantic understanding of security camera footage, making them applicable regardless of region or language. Asilla plans to integrate the outcomes of this project as the next-generation models of "AsillaVision" into its own products including "AI Security asilla," advancing deployment in the domestic market while also pursuing international expansion in collaboration with partner companies.

Comment from Masahiro Wakasa, Director & CTO, Asilla, Inc.

We are deeply honored to have our proposal selected for the "Research and Development Project for Strengthening Post-5G Information and Communication Systems Infrastructure / Development of Competitive Generative AI Foundation Models (GENIAC)" by METI and NEDO.

While the development of general-purpose foundation models is advancing on a global scale, how to build foundation models that are truly useful on the ground in each industry is a critical challenge that Japan's AI development must confront. This project aims to develop a highly practical foundation model leveraging proprietary data in the field of security — where the social challenge of labor shortage is becoming increasingly severe — and to resolve these societal issues.

We feel a great sense of responsibility for being given this opportunity, and we are committed to giving back the results and insights gained to industries across the field, while also contributing to the advancement of Japan's AI development ecosystem as a whole.

About "AI Security asilla"

AI Security asilla is a system in which AI analyzes existing security camera footage 24 hours a day, 365 days a year, instantly detecting anomalous behaviors such as violence, falls, and intrusions, as well as cautionary behaviors such as wandering, crowding, and physical distress. As the shortage of security personnel grows increasingly serious, the system captures irregularities that human monitoring tends to miss, and immediately notifies security staff and facility managers. Since it works with existing cameras without additional equipment investment, it enables high levels of safety to be maintained even with limited personnel — a next-generation security solution.

Company Overview

Asilla, Inc.Representative: Representative Director & CEO, Tsuyoshi OnoueHeadquarters: 1-4-2 Nakamachi, Machida, TokyoBusiness: Development and provision of various products and solutions based on action recognition AIOfficial website: https://jp.asilla.com/

Press Inquiries

Asilla, Inc.

Public Relations: Nakamura

Email: pr@asilla.jp

No items found.

No items found.
VIEW MORE

CONTACT US

Contact Us
For inquiries about our products, please contact us.