English 
Knowledge Communication Inc. will begin offering a generative AI security solution from today to help companies use generative AI safely and securely in their business operations as the adoption and use of generative AI accelerates.
This service combines multiple advanced technologies, such as Azure AI Content Safety, AWS Bedrock Guardrails, F5 AI Gateway, and Cisco AI Defense, to provide flexible defense configurations tailored to your needs by combining multiple security services for specific risks associated with generative AI, such as prompt injection, personal information leakage, output of false information (hallucination), malicious data, and erroneous learning (model poisoning). As the adoption and use of generative AI progresses rapidly, this service comprehensively covers new security challenges that organizations face.
こちらもお読みください： HENNGE One Partners with Cloud System “MENTENA”
Generative AI and AI agents are being rapidly adopted and used by many companies. However, security risks specific to LLM, such as the following, are suddenly emerging.
Prompt injection: The risk of AI behavior being unintentionally rewritten by unauthorized external input
Output of misinformation, prejudice, and harmful content: Possibility of damaging business judgment and brand value
Vulnerability in RAG (Retrieval-Augmented Generation) configuration: Risk of confidential data aggregated in the vector DB being targeted.
ソース PRタイムズ
日本語