Secure AI on AWS Health Check

• **AWS Security Agent:** Web application and agent spaces deployed per account with Accenture-framework-based custom security requirements and AWS-managed requirements. Security Agent remains deployed post-engagement, enabling your teams to continuously run security design reviews, code reviews, penetration tests, and threat modeling against your AI infrastructure.
• **AWS Config:** Accenture-framework-based Secure AI Config conformance pack and AWS-managed secure AI conformance packs deployed, providing real-time AI infrastructure configuration monitoring feeds for existing Security Hub CSPM deployments.
• **Remediation Roadmap Infrastructure:** S3 bucket with KMS encryption for centralized findings storage and an AgentCore agent for remediation roadmap generation (partially manual until AWS APIs are released).

### Engagement Workflow:

**Step 1 – Requirements Gathering:** Determine scope (AWS accounts, required agent spaces, repos), review AI infrastructure security requirements, gather design documentation, enable deployment team access to AWS and repository.

**Step 2 – Deployment:** Via CI/CD, create Security Agent spaces, set up user access via IAM/SSO, integrate with supported repositories, add security requirements, and deploy conformance packs.

**Step 3 – Findings Generation:** Run initial AI infrastructure design and code reviews, gather and centralize findings, evaluate via Accenture agentic tooling.

**Step 4 – Reporting:** Create aggregated AI infrastructure remediation roadmap, generate insights and recommendations, conduct readout with stakeholders, knowledge transfer on Security Agent and conformance pack usage.

### Customer Outcomes

• AWS Security Agent deployed and yours to keep—enabling continuous security design reviews, code reviews, and penetration testing of your AI infrastructure
• Accenture-framework and AWS-managed security requirements active per account
• AWS Config secure AI conformance packs deployed for persistent AI infrastructure monitoring
• Aggregated findings report with prioritized remediation roadmap
• Clear path to Foundational Guardrails and AgentCore Accelerator (2-week follow-on)

### AI Infrastructure Focus

This engagement specifically assesses the security posture of your AI infrastructure: Amazon Bedrock configurations, agent architectures, model access controls, and supporting AWS services. Assessment maps to Accenture's 10-dimension Secure AI Framework aligned with NIST AI RMF, OWASP LLM Top 10, OWASP ASI threat categories, and AWS Defense in Depth for AI.

### AWS Services Activated

Security Agent, AWS Config (conformance monitoring), Bedrock AgentCore (AI workload context), S3, IAM, CloudWatch Logs, KMS (infrastructure support)