AI Security Maturity Model·AWS v1.0.0

Drag to explore · pinch to zoom
Home / Matrix / Capability 1 / c1.10
Enhanced Logging and MonitoringDefense-in-depth layer: AI Application Capability 1 · Model inferenceUse case: AI that answers 👤 CustomerAWS Shared Responsibility Model: who implements and operates this control. Customer = you own it; AWS = AWS operates it and you inherit it; Shared = both have obligations.

c1.10 Enable AI-aware threat detection

Continuously watch for anomalous API activity and abuse patterns around your inference workloads and aggregate findings centrally.

Foundational
Enhanced
Advanced
Effort: Medium Impact: Medium GenAI scopesFrom the AWS Generative AI Security Scoping Matrix.: 3 · Pre-trained models

Implementation guidance

How to implement on AWS

Enable Amazon GuardDuty to detect anomalous API and credential activity, and aggregate findings in AWS Security Hub. Use GuardDuty Malware Protection for S3 to scan user-submitted files that AI workloads process.

Acceptance criteria

GuardDuty is enabled and reporting to Security Hub; malware scanning covers user-upload buckets; a sample finding triggers your response workflow.

Practical implementation

How this control is implemented in each reference architecture:

GuardDuty runs org-wide from the delegated admin in Security tooling, watching the Generative AI and Application accounts and aggregating findings in Security Hub.

  1. Enable Amazon GuardDuty across the organization from the delegated administrator in the Security tooling account, covering the Generative AI and Application accounts.
    Security tooling GuardDuty AWS Organizations
  2. Turn on GuardDuty Malware Protection for S3 on buckets that receive user-submitted files the AI workload processes.
    Generative AI account GuardDuty Amazon S3
  3. Aggregate findings in AWS Security Hub and route high-severity ones to your response workflow via EventBridge.
    Security tooling Security Hub

How to validate: GuardDuty is enabled org-wide and reporting to Security Hub; malware scanning covers user-upload buckets; a sample finding triggers the response workflow.

Avoid: Enabling GuardDuty in only one account. Turn it on org-wide from the delegated admin so the AI accounts are covered and findings land in one place.

Generative AI OU (multi-account) →

GuardDuty and Security Hub run in the one account, with malware scanning on user-upload buckets and findings wired to a response workflow.

  1. Enable Amazon GuardDuty in the account to detect anomalous API and credential activity.
    In-account logging GuardDuty
  2. Turn on GuardDuty Malware Protection for S3 on buckets that hold user-submitted files.
    In-account logging GuardDuty Amazon S3
  3. Aggregate findings in Security Hub and alert on high-severity ones.
    In-account logging Security Hub

How to validate: GuardDuty reports to Security Hub; malware scanning covers upload buckets; a sample finding triggers your alert.

Avoid: Deferring threat detection until 'later'. Enable GuardDuty from the start; it is agentless and low-cost, and the AI workload is an attractive target.

Single account (entry point) →

Threats addressed

  • Abnormal invocation or credential-use patterns going unnoticed.
  • Resource-abuse or exfiltration attempts across accounts.

AWS services

Amazon GuardDuty AWS Security Hub

Framework mappings

OWASP GenAI/LLM Top 10 (2026)

  • LLM02 Sensitive Information Disclosure
  • LLM06 Unbounded Consumption

NIST AI Risk Management Framework (AI 100-1)

  • MEASURE.3 Mechanisms for tracking AI risks are in place
  • MANAGE.4 Risk treatments, monitoring and response are documented and applied

AWS Well-Architected Generative AI Lens (Security)

  • GENSEC03-BP01 Implement control plane and data access monitoring

Learning resources

References

  • AWS SRA for AI — Capability 1, AI inference account