日本語 
ProVision Co., Ltd. rolled out ‘AI System Vulnerability Quick Diagnosis.’ It is meant to spot security risks in AI setups in about one week.
The review looks at weak spots that may show up from many kinds of inputs. These include text, pictures, audio, and uploaded documents.
The service checks for risks such as prompt injection, confidential information leakage, unintended AI behavior, and vulnerabilities linked to RAG systems. It combines a common assessment with additional checks based on how the target AI system receives and processes information.
Also Read: UNDP and NEC Sign Memorandum of Understanding on the Use of AI and Other Digital Technologies to Realize a Nature-Positive Society
For text systems, testing includes multilingual and encoding-based attacks, indirect prompt injection, and misinformation. Image, audio, and document-based systems are assessed for embedded instructions, excessive automated actions, and confidential data exposure.
The diagnostic approach is based on the OWASP Top 10 for LLM Applications. The launch reflects a growing need for AI-specific security testing as enterprises move beyond traditional chatbots toward RAG, multimodal AI, and AI agents.
English