# Does each AI agent, pipeline and application have its own cloud identity, not a shared one or a person's credentials?

Facts read 16 Nov 2026. Page reviewed 16 Nov 2026.

Question 23 of the assessment asks: Does each AI agent, pipeline and application have its own cloud identity, not a shared one or a person's credentials? If an agent shares an identity, you cannot tell what it did and you cannot cut its access without breaking other things.

## Why it matters

If an agent shares an identity, you cannot tell what it did and you cannot cut its access without breaking other things.

## Answers and what each is worth

| Answer | Value |
| --- | --- |
| Yes, one identity per agent or application | 1 |
| Some share an identity | 0.5 |
| No or not known | 0 |

## How to check it in each cloud

- **AWS:** Check that each Bedrock agent, Lambda or container that calls models has its own execution role, and that CloudTrail userIdentity shows distinct roles per agent.
- **Azure:** Check that each AI app uses its own managed identity and that Azure OpenAI or AI Foundry role assignments are not made to a shared principal.
- **Google Cloud:** Check that each Vertex AI agent or job runs as its own service account, not the default Compute Engine service account.

## The fix

Give every AI agent and pipeline a dedicated identity and remove the shared roles and personal tokens it used. Effort: medium.

## Where it comes from

- Microsoft cloud security benchmark AI-4: [Apply least privilege for agent functions](https://learn.microsoft.com/en-us/security/benchmark/azure/mcsb-v2-artificial-intelligence-security)
- NIST SP 800-53 Rev. 5 IA-2: [source](https://csrc.nist.gov/pubs/sp/800/53/r5/upd1/final)

## Related pages

- [Question 24: AI workloads, ai tool permissions](/questions/ai-tool-permissions/)
- [Question 25: AI workloads, approved models only](/questions/approved-models-only/)
- [Question 26: AI workloads, ai data reach](/questions/ai-data-reach/)

## Sources

- Microsoft cloud security benchmark AI-4: https://learn.microsoft.com/en-us/security/benchmark/azure/mcsb-v2-artificial-intelligence-security (read 16 Nov 2026)
- NIST SP 800-53 Rev. 5 IA-2: https://csrc.nist.gov/pubs/sp/800/53/r5/upd1/final (read 16 Nov 2026)

Last reviewed 16 Nov 2026. Canonical: https://cloudsecurityriskassessment.com/questions/ai-own-identity/
