Bedrock Guardrails Won-t Stop an Exfiltrating Agent, But Here's What Will

If a compromised Bedrock agent tried to exfiltrate data from your environment, Bedrock Guardrails won’t stop it. Explore a containment architecture that will.  

Govern AWS Bedrock Agents With Default-Deny Enforcement 

Bedrock Guardrails Won't Stop an Exfiltrating Agent, But Here's What Will

451 Research Report: Aviatrix extends cloud security fabric to AI agents with containment platform launch

A 451 Research report by Mike Fratto highlights Aviatrix's agentless containment: cloud security architecture built for rapidly-changing agentic threats.

451 Research Report: Aviatrix's Agentless Containment

451 Research Report Profiles Aviatrix Containment Platform

Lateral movement in cloud security is when an attacker who has gained initial access to one resource moves to other resources within the cloud environment. They typically do this using stolen credentials, valid service accounts, or open east-west network paths between workloads. The goal is to escalate privileges, access sensitive data, and expand control before detection.


What is lateral movement in cloud security? 

In on-premises environments, lateral movement typically happens across a flat internal network. In cloud environments, attackers move across VPCs, accounts, regions, and cloud providers using Application Programming Interface (API) calls, Identity and Access Management (IAM) roles, and service account credentials. Cloud workloads are also ephemeral, which means traditional Internet Protocol(IP)-based segmentation often does not apply consistently.


How is lateral movement in the cloud different from lateral movement in on-premises networks? 

A traditional perimeter firewall cannot prevent cloud lateral movement on its own. Firewalls govern north-south traffic entering and leaving the network, but east-west traffic between workloads within the cloud rarely passes through a perimeter firewall. Preventing lateral movement requires inline enforcement at the workload level for east-west traffic.


Can a firewall prevent lateral movement in the cloud? 

Microsegmentation divides the cloud environment into fine-grained trust zones and restricts communication between them. When implemented with workload identity-based policies rather than IP rules, microsegmentation ensures that a compromised workload cannot reach adjacent resources without explicit policy authorization. This directly limits the Blast Radius of any lateral movement attempt.

A more comprehensive solution than microsegmentation is [Communication Governance](https://aviatrix.ai/learn-center/communication-governance/), an enforcement model that governs every communication path between workloads in a cloud environment with a default-deny policy.


What is the role of microsegmentation in preventing lateral movement in the cloud?

Zero Trust prevents lateral movement by removing implicit trust between workloads and requiring continuous verification of every connection. Workload-to-workload communication is deny-by-default. Every session must match an authorized policy based on workload identity, context, and least-privilege access rules. Even if an attacker holds valid credentials, they cannot access resources without policy authorization.


How does Zero Trust prevent lateral movement in the cloud? 

The primary MITRE ATT&CK techniques for cloud lateral movement include TA0008 (Lateral Movement tactic), T1021 (Remote Services), T1078 (Valid Accounts), and T1570 (Lateral Tool Transfer). All of these techniques depend on east-west network reachability between cloud workloads. Removing those paths through deny-by-default segmentation directly constrains these techniques.


What MITRE ATT&CK techniques are used for cloud lateral movement? 

Prevent Lateral Movement Before Attackers Gain Footholds

Learn how to prevent lateral movement in cloud environments with Zero Trust segmentation, identity-based policies, and runtime enforcement before attackers escalate a breach.

How to Prevent Lateral Movement in the Cloud

Practical Takeaways from NSA’s Zero Trust Guidance

The NSA Zero Trust Implementation Guidelines are extensive and specific. Read an actionable summary to start your Zero Trust journey for control and visibility. 

The TL;DR on NSA’s Zero Trust Implementation Guidelines

Woman studying a tablet against a background of screens.

Stopping Shadow AI at the Network Layer card image

Stopping Shadow AI at the Network Layer

AI agents are powerful tools, but shadow AI threatens security and compliance—and cloud-native tools fall short. Learn how zero trust provides governance. 

Govern Shadow AI with Zero Trust Security

Controlled vs. Autonomous The Two AI Security Challenges Reshaping Your Tech Stack 2 card image

Controlled vs. Autonomous The Two AI Security Challenges Reshaping Your Tech Stack 2

Human-prompted AI and autonomous agents create unexpected risks. Learn how to create defense in depth and translate compliance requirements into action.

Controlled vs. Autonomous: Two Security Risks To Watch

Controlled vs. Autonomous: The Two AI Security Challenges Reshaping Your Tech Stack

The Hidden Risk in Your Cloud What’s Really Happening Between Your Cloud Workloads

The Hidden Risk in Your Cloud What’s Really Happening Between Your Cloud Workloads v4

Traditional perimeter-based security has huge blind spots that allow lateral movement. Learn how to regain control of your network through full visibility.

The Hidden Risk in Your Cloud: What’s Really Happening Between Your Cloud Workloads?

IBM 2025 Cost of Data Breach Report Reveals Shadow AI Risks card image

IBM 2025 Cost of Data Breach Report Reveals Shadow AI Risks

IBM’s 2025 report finds AI model/app breaches rising. Learn why Shadow AI evades legacy tools and how a Cloud-Native Security Fabric secures enterprise AI.

Sam Pandey

Author Biography

Published Articles by Sam Pandey

Bedrock Guardrails Won't Stop an Exfiltrating Agent, But Here's What Will

451 Research Report Profiles Aviatrix Containment Platform

Prevent Lateral Movement Before Attackers Gain Footholds

Practical Takeaways from NSA’s Zero Trust Guidance

Stopping Shadow AI at the Network Layer

Controlled vs. Autonomous: The Two AI Security Challenges Reshaping Your Tech Stack

The Hidden Risk in Your Cloud: What’s Really Happening Between Your Cloud Workloads?

IBM 2025 Cost of Data Breach Report Reveals Shadow AI Risks

Products

By Role

Customer Stories

Security & Legal

Solutions

Resources

Industry

Company

Tools