#Japan #Tokyo #AI Safety #SherLOCK #AISI

SherLOCK's Contribution to AI Safety in Healthcare

SherLOCK Inc., headquartered in Minato, Tokyo, is making waves in the field of AI security by collaborating with the AI Safety Institute (AISI). As a member of AISI's healthcare sub-working group, SherLOCK has played a vital role in developing the "AI Safety Evaluation Perspective Guide" tailored for the healthcare sector. This guide aims to equip companies with practical and effective evaluation methods to ensure safety and quality in AI applications.

Background and Objectives

The integration of generative AI into healthcare is expected to bring about significant innovations. However, challenges such as hallucinations (the generation of inaccurate information), privacy concerns, and security issues are prevalent. The first Hiroshima Global Forum for Trustworthy AI, held in January 2026, highlighted the international importance of achieving "Trustworthy AI." Recognizing the urgent need for reliable AI, SherLOCK has participated in this initiative, leveraging its technical expertise to assist businesses in establishing safe operational environments that balance business value with consumer safety and confidence.

Key Areas of Support by SherLOCK

During the formulation of the AI Safety Evaluation Guide, SherLOCK contributed significantly in several areas:

• - Security and Data Quality Assessment: SherLOCK provided expert knowledge on evaluation criteria and points related to AI red teaming tests, which assess security and data quality. Considerations for risks such as generating misleading information through malicious prompts or facilitating data leaks were included, allowing for preemptive measures against critical incidents in healthcare environments.
• - Data Quality Assessment for RAG: The guide delineates the risks related to data quality used in technologies like Retrieval-Augmented Generation (RAG), ensuring the accuracy of medical and health-related information.
• - Advice on Third-Party Evaluations: SherLOCK also provided insights on how third-party evaluations, particularly through AI red teaming tests, can bolster a company's credibility and trustworthiness in the realm of AI safety and security.

Features of the Guide

The AI Safety Evaluation Guide is crafted to be accessible for companies that may lack deep expertise in the field:

• - Five-Phase Evaluation Aligned with AI Lifecycle: Each phase, including product design, model selection, implementation, verification, and operation, is outlined with evaluation points clearly specified.
• - Ten Multi-Faceted Safety Evaluation Perspectives: The guide encompasses a wide range of safety evaluation aspects, from universal criteria to sector-specific risks, highlighting tangible risks—such as direct harm to patient life and health—if evaluations are neglected.
• - Practical Risk Management Support: It features organized assessments of various risks and countermeasures, enabling on-ground staff to apply the guidelines immediately without extensive additional training.

Accessing the Guide

The AI Safety Evaluation Perspective Guide for the healthcare sector can be downloaded from the following link: AISI Guide Download

Contributions from SWG Members

AISI's healthcare sub-working group includes industry leaders such as Ubie, Awarefy, Ajinomoto, SB Intuitions, and other prominent healthcare organizations, collaborating to enhance AI’s safe implementation in healthcare settings.

Comments from CEO Teresa Tsukiji

Teresa Tsukiji, CEO of SherLOCK, expressed pride in contributing to the foundational AI Safety Evaluation Guide alongside AISI and SWG members, emphasizing the essential role of AI safety assessments in facilitating safe innovations rather than hindering them. She reassured that SherLOCK is committed to supporting the safe implementation of context-sensitive AI across multiple domains, including healthcare, through cutting-edge AI security technologies and governance support.

About SherLOCK Inc.

• - Company Name: SherLOCK Inc.
• - CEO: Teresa Tsukiji
• - Founded: January 2024
• - Location:5-9-1 Toranomon, Minato, Tokyo, Japan
• - Business Focus: Development and sale of AI security/safety solutions, enhancement of AI governance, and related R&D support.
• - Website: SherLOCK Official Website