Rethinking AI Security to Deliver Better Outcomes - Check Point Software

‍

Overview

In this masterclass from the CISO and CIO Inspired Summit at Carton House, Ireland, Eduardo Redondo Garcia, Global Head of Cloud Security Architecture at Check Point Software Technologies, shares part of Check Point's strategy for protecting generative AI, one of the most transformative technologies in decades.

As generative and agentic AI transform customer experiences and expectations, Eduardo explains the new risks they introduce, from data leakage and model abuse to prompt injection, and how to augment traditional security approaches and integrate AI security with cloud-native services.

Key takeaways

  • AI is not a passing trend; it is transforming how we live, work and do business, and security strategies need to keep pace.
  • Employees can unintentionally leak tokens, secrets, passwords and personal data into public AI tools, for example when reviewing a contract.
  • Simply blocking tools can push users to bypass controls with other tools, so guidance and visibility matter as well as enforcement.
  • Risks include misuse of access controls by agents, model abuse and prompt injections that existing tools will likely miss, aimed at malicious output or data exfiltration.
  • Red teaming your language models helps validate how well they are protected, and games such as Lakera's Gandalf show how easily prompts can extract secrets.

Watch the full session on YouTube: Rethinking AI Security to Deliver Better Outcomes | Eduardo Redondo Garcia, Check Point Software