LATEST INTELLIGENCE

PRESENTED BY

Securing GenAI

A COMPREHENSIVE REPORT ON PROMPT ATTACKS: TAXONOMY, RISKS, AND SOLUTIONS

Generative AI( GenAI) has seen a remarkable surge in popularity, transforming productivity across a wide range of sectors and everyday tasks. However, this rapid adoption has also introduced significant security challenges. What new risks and attack vectors have emerged? How severe are they? And can traditional security solutions effectively safeguard the use of AI?

Mainstream LLMs vs Prompt-Based Attacks

Whether employees access AI, enterprise AI-based applications, or AI agents, prompt attacks are the fundamental threats. Adversarial prompt attacks manipulate GenAI systems by crafting deceptive inputs, resulting in unintended or harmful outputs. While various efforts have been made to categorize these attacks, creating a comprehensive taxonomy remains a challenge. Existing classifications often fail to keep pace with new attack vectors, making it difficult to adapt or map evolving threats to predefined categories.

An Impact-Focused Taxonomy

Download whitepaper here

We recently assessed mainstream large language models( LLMs) against prompt-based attacks, which revealed significant vulnerabilities. Three attack vectors – guardrail bypass, information leakage, and goal hijacking – demonstrated consistently high success rates across various models. In particular, some attack techniques achieved success rates exceeding 50 % across models of different scales, from several-billion- parameter models to trillion-parameter models, with certain cases reaching up to 88 %.

To address these gaps, this whitepaper proposes a comprehensive, impact-focused taxonomy for adversarial prompt attacks. It provides a detailed mapping of existing AI attack techniques within this taxonomy, shedding light on their potential consequences and impact on the application and implementation technique. p

Download whitepapers free from www. intelligentcio. com / eu / whitepapers /

20 INTELLIGENTCIO EUROPE www. intelligentcio. com