Risky Bulletin Podcast – Episode Summary

Episode Title: Sponsored: What AI workloads mean for Cloud security
Date: January 11, 2026
Host: Tom Uren
Guest: Tony De la Fuente, CEO and Founder of Prowler

Overview

This episode explores how the rapid growth of artificial intelligence (AI) workloads is reshaping cloud security, featuring insights from Tony De la Fuente, founder of Prowler—an open source cloud native security company. Tom and Tony discuss cloud’s emergence as the fundamental layer of applications, the intertwined evolution of AI and cloud infrastructure, challenges in securing AI platforms, the use of “attack path” analysis, and the current state of knowledge in the industry.

Key Discussion Points & Insights

The Cloud as the New Operating System

• Tony likens the cloud to the operating system of the modern era:

  “The cloud is the new operating system.” [01:21 – Tony]

• Applications are now built in the cloud (AWS, Azure, GCP) instead of on traditional OSes.
• Prowler was designed to secure this “new operating system,” automating security and compliance tasks.

AI & Cloud: Symbiotic Relationship

• AI workloads are deeply dependent on the cloud for scalable computation, data storage, and APIs.

  “AI and cloud are kind of family members, right? Because AI needs the cloud and the cloud uses AI for everything.” [01:50 – Tony]

• Protection requirements for both AI and traditional cloud workloads have significant overlap, but AI introduces additional services/components to consider.

Securing AI Services in the Cloud

• Security focus extends beyond the core AI service (e.g., AWS Bedrock, SageMaker) to adjacent infrastructure:
  • API endpoints/gateways
  • Storage (e.g., S3 buckets)
  • Databases
  • Service configurations/permissions

  “When it comes to securing cloud services for AI…you have to take into account not only that service itself, but all the services around it.” [02:43 – Tony]

• Importance of least privilege for API keys and tight access control.
• Need to avoid sensitive data exposure and ensure proper guardrails.

AI Workloads and New Security Challenges

• While many underlying services are familiar (e.g., S3, API gateways), AI stacks introduce new architectural complexity.

• Securing AI architectures means considering:

  • Tool and data integrations
  • API token management
  • Data flow and privilege boundaries

  “If you are building an application with AI…first of all, you need to make sure your users are safe…to prevent jailbreaks, to prevent data leaks, to prevent business violations.” [07:23 – Tony]

• Concrete risks:

  • Prompt injection
  • Information leakage
  • Inappropriate model/tool access (e.g., SageMaker notebooks with root exposure)
  • Overprivileged accounts easily exploited by attackers

Multi-cloud Trends with AI

• Most organizations use different clouds for different workloads (rather than using the same type of AI workload for redundancy across clouds).

  “A trend that we see is using multiple clouds for different things…not for exactly the same.” [04:47 – Tony]

• Large enterprises may sometimes build redundancy this way, but it’s uncommon.

AI Model Security and Governance

• Security doesn't stop at infrastructure—audit and monitor the AI models themselves.
• Prowler’s recommendations include:
  • Scanning models for vulnerabilities
  • Ensuring guardrails (e.g., prompt injection prevention)
  • Monitoring logs for suspicious invocation or abuse
  • Following new standards like OWASP for AI

  “The basics are probably the OWASP, the new OWASP for AI, but beyond that.” [06:47 – Tony]

State of Industry Knowledge

• Security professionals are still learning AI/cloud security best practices.

  “I think we are in very early days for all this stuff. So that is why tools like Prowler or other tools…are very important for the industry.” [10:01 – Tony]

• Emphasized importance of covering “the basics,” especially in proofs-of-concept or early deployments.

Lessons from the Early Web

• Tony draws parallels to the evolution from static HTML to dynamic web apps: as complexity grows, so do attack surfaces and the need for layered security.

  “Adding the AI component into a workload adds all of that with its own pros and cons.” [11:23 – Tony]

Notable Quotes & Memorable Moments

• “The cloud is the new operating system.”
  [01:21 – Tony De la Fuente]
• On AI’s relationship to cloud:
  “AI and cloud are kind of family members, right? Because AI needs the cloud and the cloud uses AI for everything.”
  [01:50 – Tony De la Fuente]
• “We are in very early days for all this stuff. That is why tools like Prowler or other open source tools, are very important…”
  [10:01 – Tony De la Fuente]
• “A good example can be Amazon SageMaker…If you don't configure properly those SageMaker notebooks you can expose those notebooks to the Internet and an attacker can use those notebooks. And... you can set up root access to those notebooks. So it's a perfect recipe for an attack.”
  [08:44 – Tony De la Fuente]

Attack Path: Visualizing Cloud Risk

• “Attack path” is Prowler’s new feature for visualizing risks.

  “Attack path in the cloud means…to see how all the components connect each other. How to connect the dots from the Internet, let's say…to the data.”
  [13:00 – Tony De la Fuente]

• Explains how in cloud environments, attackers may chain together vulnerabilities, misconfigurations, and excessive permissions to reach sensitive data.
• Attack paths are more complex in cloud vs. on-premises due to dynamic roles, APIs, and interconnected services.
• Practical example:
  1. VM/EC2 instance with vulnerable application
  2. Instance has overprivileged access to S3 bucket with PII
  3. Attacker exploits app, pivots through permissions, accesses sensitive data

Timestamps for Key Segments

• 00:03–01:03: Introductions, Prowler’s mission in securing the “cloud as OS”
• 01:03–02:43: How AI and cloud security are intertwined
• 02:43–03:28: Adjusting security approaches for new AI-related threats and architectures
• 04:26–05:44: Trends in multi-cloud AI platform adoption and redundancy strategies
• 06:19–07:23: Securing AI models and workloads—guardrails, prompt injection, OWASP AI
• 07:18–09:43: Real-world risks/infrastructure vulnerabilities (e.g. SageMaker notebooks)
• 10:01–11:23: State of practitioner knowledge and learning from web security history
• 12:10–15:00: “Attack path” visualization—explaining risk chains in cloud environments

Takeaways

• The shift to cloud as the application “OS” means security leaders must expand their focus beyond traditional servers and network boundaries.
• AI workloads inherit old risks, gain novel ones, and compound the complexity of modern attack surfaces.
• Visualization tools (like Prowler's "attack path") help teams grasp and mitigate multi-stage, cloud-native threats.
• The industry is early on the AI/cloud security journey; prioritizing the basics, open source tooling, and insight sharing is essential for raising the security bar.

Guest: Tony De la Fuente, CEO and founder of Prowler
Host: Tom Uren
Podcast: Risky Bulletin

(Advertisements, intros, and outros were excluded from this summary.)