GenAI workload security gaps: what IAM teams need to know

TL;DR: Forty-one percent of organisations say they are struggling with security risks and vulnerabilities when integrating GenAI into their AI infrastructure, according to Cyera's report with ESG and AWS. The real issue is that GenAI workloads now handle sensitive data and critical workflows faster than existing identity and data controls can reliably govern.

NHIMG editorial — based on content published by Cyera: Importance of Securing Workloads for Generative AI Report AI

By the numbers:

• 41% of organisations report struggling with security risks and vulnerabilities when integrating GenAI into their AI infrastructure.

Questions worth separating out

Q: How should security teams govern GenAI workloads that access sensitive data?

A: Treat the workload as a governed identity with explicit scope, named ownership, and continuous review.

Q: Why do GenAI integrations create security risk even when the model is approved?

A: Approval of the model does not guarantee control over the runtime environment.

Q: What do IAM and security teams get wrong about GenAI access control?

A: They often focus on the model or the application wrapper and ignore the data paths behind it.

Practitioner guidance

• Classify each GenAI workload as a governed identity Inventory the model, orchestration layer, connectors, and downstream data paths as one access chain.
• Map sensitive data paths through prompts and connectors Trace where confidential data enters, is transformed, cached, exported, or logged.
• Tighten workload access to task-scoped privileges Remove broad read or write permissions that are not required for the current use case.

What's in the full article

Cyera's full report covers the operational detail this post intentionally leaves for the source:

• How Cyera, ESG, and AWS break down the workload-security findings behind GenAI adoption pressure
• The report's supporting survey context around security risk, compliance, and business continuity concerns
• Implementation framing for protecting AI assets and sensitive data across production GenAI environments

👉 Read Cyera's report on securing workloads for generative AI →

GenAI workload security gaps: what IAM teams need to know?

Explore further

View Full Forum →  |  NHI Foundation Course →