The AI Security Gap: Why Protecting Prompts Isn’t Enough | CSA Summit 2025 at RSAC
Released: 04/29/2025
Updated: 11/17/2025
LLM firewalls play a crucial role in filtering prompts and responses, but they only address a fraction of AI security risks. Protecting AI systems requires a comprehensive, system-level approach that mitigates threats across the entire AI pipeline—from data ingestion to model deployment and consumption. The 2025 OWASP Top 10 for LLMs provides a critical framework for understanding and addressing AI security vulnerabilities, moving beyond the misconception that securing GenAI is only about protecting prompts or models.
Explored in this session was:
- Why AI security must extend beyond prompt-level protection to full system security
- How the OWASP Top 10 for LLMs defines AI system-level risks and mitigation strategies
- Essential security controls for every stage of the AI pipeline lifecycle
- How to effectively leverage a data-context aware LLM Firewall
- How to implement a robust security strategy aligned with OWASP’s AI security framework
Gain actionable insights on securing AI systems holistically and ensuring resilient, enterprise-grade AI adoption.
Presented by:
Presented by:
- Rehan Jalil, CEO, Securiti
- Rex Thexton, CTA, Accenture Security
Download this Resource
