Simuna InfosecSIMUNA INFOSEC
AI Security2026-08-05

Keamanan AI & LLM: Panduan OWASP Top 10 for LLM Applications 2025

Aplikasi AI memperkenalkan risiko keamanan baru. Panduan pengujian berdasarkan OWASP Top 10 for LLM Applications 2025 untuk perusahaan Indonesia.

As Indonesian enterprises deploy AI applications across e-commerce, fintech, and banking, the OWASP Top 10 for LLM Applications 2025 provides the framework for testing these AI-specific risks. Traditional security tools cannot detect prompt injection, data leakage, or excessive agency — risks unique to AI systems.

The 2025 edition covers ten categories including Prompt Injection (the number-one risk), Sensitive Information Disclosure, Data and Model Poisoning, Excessive Agency, and two new categories: System Prompt Leakage and Vector and Embedding Weaknesses (targeting RAG pipelines).

For OJK-regulated financial institutions deploying AI, these risks fall within the scope of scenario-based testing that POJK 11/2022 and SEOJK 29 require. AI applications processing personal data must also be assessed under the now-enforceable UU PDP (Law 27/2022).

Our AI/LLM VAPT service tests against the complete OWASP Top 10 for LLM Applications 2025, combining traditional application security with AI-specific attack techniques. We have experience securing a major Indonesian telecom operator and bring this enterprise rigour to AI security testing.

*This article references the OWASP Top 10 for LLM Applications 2025.*