DiscoverAWS for Software Companies PodcastEp150: Security Considerations for Generative AI with CyberArk, Fortra and Sysdig
Ep150: Security Considerations for Generative AI with CyberArk, Fortra and Sysdig

Ep150: Security Considerations for Generative AI with CyberArk, Fortra and Sysdig

Update: 2025-09-26
Share

Description

Security leaders from CyberArk, Fortra, and Sysdig share actionable strategies for securely implementing generative AI and reveal real-world insights on data protection and agent management.

Topics Include:

  • Panel explores practical security approaches for GenAI from prototype to production
  • Three-phase framework discussed: planning, pre-production, and production security considerations
  • Security must be built-in from start - data foundation is critical
  • Understanding data location, usage, transformation, and regulatory requirements is essential
  • Fortra's security conglomerate approach integrates with AWS native tools and partners
  • Machine data initially easier for compliance - no PII or HIPAA concerns
  • Identity paradigm shift: agents can dynamically take human and non-human roles
  • 97% of organizations using AI tools lack identity and access policies
  • Security responsibility increases as you move up the customization stack
  • OWASP Top 10 for GenAI addresses prompt injection and data poisoning
  • Rigorous model testing including adversarial attacks before deployment is crucial
  • Sysdig spent 6-9 months stress testing their agent before production release
  • Tension exists between moving fast and implementing proper security controls
  • Different security approaches needed based on data sensitivity and model usage
  • Zero-standing privilege and intent-based policies critical for agent management
  • Multi-agent systems create "Internet of Agents" with exponentially multiplying risks
  • Discovery challenge: finding where GenAI is running across enterprise environments
  • API security and gateway protection becoming critical with acceptable latency
  • Top customer need: translating written AI policies into actionable controls
  • Threat modeling should focus on impact rather than just vulnerability severity


Participants:


Further Links:


See how Amazon Web Services gives you the freedom to migrate, innovate, and scale your software company at https://aws.amazon.com/isv/

Comments 
In Channel
loading
00:00
00:00
1.0x

0.5x

0.8x

1.0x

1.25x

1.5x

2.0x

3.0x

Sleep Timer

Off

End of Episode

5 Minutes

10 Minutes

15 Minutes

30 Minutes

45 Minutes

60 Minutes

120 Minutes

Ep150: Security Considerations for Generative AI with CyberArk, Fortra and Sysdig

Ep150: Security Considerations for Generative AI with CyberArk, Fortra and Sysdig