> [!NOTE]
> Premium and Enterprise (Deprecated) [service tiers](https://docs.cloud.google.com/security-command-center/docs/service-tiers.md.txt)

> [!CAUTION]
> The Security Command Center Enterprise service tier will shut down on May 21, 2027. If your organization uses the Enterprise service tier, then the organization will automatically move to the Premium service tier on or after May 21, 2027. Contact your account representative or a [Google Cloud sales specialist](https://cloud.google.com/contact) if you have questions.

This document offers informal guidance on how you can respond to findings of suspicious
activities in your AI resources. The recommended steps might not be appropriate for all
findings and might impact your operations. Before you take any action, you should investigate the
findings; assess the information that you gather; and decide how to respond.

> [!IMPORTANT]
> **Note:** The techniques that Security Command Center provides aren't guaranteed to be effective against any previous, current, or future threats that you face. For information about why Security Command Center does not provide official remediation guidance for threats, see [Approach to threat finding remediation](https://docs.cloud.google.com/security-command-center/docs/how-to-investigate-threats.md.txt#remediating_threats).

## Before you begin

1. [Review the
   finding](https://docs.cloud.google.com/security-command-center/docs/how-to-investigate-threats#reviewing_findings). Note the affected resources and the detected binaries, processes, or libraries.
2. To learn more about the finding that you're investigating, search for the finding in the [Threat findings
   index](https://docs.cloud.google.com/security-command-center/docs/threat-findings-index).

## General recommendations

- Contact the owner of the affected resource.
- Work with your security team to identify unfamiliar resources, including Agent Runtime instances, sessions, service accounts, and [agent identities](https://docs.cloud.google.com/agent-builder/agent-engine/agent-identity). Delete resources that were created with unauthorized accounts.
- To identify and fix overly permissive roles, use [IAM
  Recommender](https://docs.cloud.google.com/iam/docs/recommender-overview). Delete or disable potentially compromised accounts.
- For the Standard-legacy, Standard, Premium, and Enterprise (Deprecated) service tiers, investigate any [identity and access
  findings](https://docs.cloud.google.com/security-command-center/docs/ciem-identity-access-findings#ciem-findings).
- For further investigation, you can use incident response services, for example, [Mandiant](https://cloud.google.com/security/consulting/mandiant-incident-response-services).
- For forensic analysis, collect and back up the logs of affected resources.

## Potentially compromised service account or agent identity

- To remove a compromised agent identity, [delete its corresponding
  Agent Runtime
  instance](https://docs.cloud.google.com/agent-builder/agent-engine/manage/overview#delete).

- [Disable the potentially compromised service
  account](https://docs.cloud.google.com/iam/docs/service-accounts-disable-enable#disabling). For best
  practices, see [Disable unused service accounts before deleting
  them](https://docs.cloud.google.com/iam/docs/best-practices-service-accounts#disable-before-delete).

- [Disable service account keys](https://docs.cloud.google.com/iam/docs/keys-disable-enable.md.txt#disabling) for
  the potentially compromised project.

- To see when your service accounts and keys were last used to call a Google
  API, use Activity Analyzer. For more information, see [View recent usage for
  service accounts and
  keys](https://docs.cloud.google.com/policy-intelligence/docs/activity-analyzer-service-account-authentication).

- If you are confident that it is safe to delete the service account,
  [delete](https://docs.cloud.google.com/iam/docs/service-accounts-delete-undelete.md.txt#deleting) it.

  > [!NOTE]
  > **Note:** If the compromised service account is a Agent Runtime [service agent](https://docs.cloud.google.com/iam/docs/service-account-types.md.txt#service-agents), you can't directly delete it. In this case, ensure that the service agent doesn't have more permissions than it needs.

- To use organization policies to restrict service account usage, see
  [Restricting service account
  usage](https://docs.cloud.google.com/resource-manager/docs/organization-policy/restricting-service-accounts).

- To use Identity and Access Management to restrict service account or service account key usage,
  see [Deny access to resources](https://docs.cloud.google.com/iam/docs/deny-access.md.txt).

## Exfiltration and extraction

- [Revoke
  roles](https://docs.cloud.google.com/iam/docs/granting-changing-revoking-access#revoking-console) for the principal that is listed in the **Principal email** row in the finding details until the investigation is complete.
- To stop further exfiltration, add restrictive IAM policies to the affected resources.
- To determine whether the affected datasets have sensitive information, [inspect them with
  Sensitive Data Protection](https://docs.cloud.google.com/sensitive-data-protection/docs/inspecting-storage). You can configure the inspection job to send the results to Security Command Center. Depending on the quantity of information, Sensitive Data Protection costs can be significant. Follow best practices for [keeping
  Sensitive Data Protection costs under
  control](https://docs.cloud.google.com/sensitive-data-protection/docs/best-practices-costs).
- Use [VPC Service Controls](https://docs.cloud.google.com/vpc-service-controls/docs/overview.md.txt) to create security perimeters around data services like BigQuery and Cloud SQL to prevent data transfers to projects outside the perimeter.

## Suspicious token generation

- Validate the need for cross-project token generation. If it's unnecessary,
  remove the IAM role binding in the target project that grants
  the `iam.serviceAccounts.getAccessToken`,
  `iam.serviceAccounts.getOpenIdToken`,
  `iam.serviceAccounts.implicitDelegation`, or
  `iam.serviceAccounts.signJwt` permission to the principal from the
  source project.

- Investigate the logs that are specified in the finding to validate the
  token generation methods used by your agentic workloads.

## Sensitive IAM roles or permissions granted

- Revoke unauthorized IAM role bindings granted to external principals on Agent Platform or Gemini Enterprise resources.
- Use Organization Policy Service to [restrict identities with domain-restricted
  sharing](https://docs.cloud.google.com/organization-policy/restrict-domains).
- To identify and fix overly permissive roles, use [IAM
  Recommender](https://docs.cloud.google.com/iam/docs/recommender-overview).

## Unauthorized Gemini Enterprise document imports

- Review the `ImportDocuments` request details in Logging to identify the target Gemini Enterprise datastore and the external principal or external source resource (such as an external Cloud Storage bucket or BigQuery dataset).
- [Revoke IAM roles](https://docs.cloud.google.com/iam/docs/granting-changing-revoking-access.md.txt#revoking-console) for the principal that initiated the unauthorized import until your investigation is complete.
- Inspect and purge any untrusted documents imported into the affected Gemini Enterprise datastore to prevent knowledge store poisoning or indirect prompt injection.
- Use [VPC Service Controls](https://docs.cloud.google.com/vpc-service-controls/docs/overview.md.txt) to enforce service perimeters around Gemini Enterprise and your data storage services to block unauthorized cross-perimeter data imports.

## What's next

- Learn [how to work with threat
  findings in Security Command Center](https://docs.cloud.google.com/security-command-center/docs/how-to-investigate-threats).
- Refer to the [Threat findings index](https://docs.cloud.google.com/security-command-center/docs/threat-findings-index.md.txt).
- Learn how to [review a
  finding](https://docs.cloud.google.com/security-command-center/docs/how-to-investigate-threats#reviewing_findings) through the Google Cloud console.
- Learn about the [services that
  generate threat findings](https://docs.cloud.google.com/security-command-center/docs/concepts-security-sources#threats).

<!-- -->

- Refer to a [list of all AI findings](https://docs.cloud.google.com/security-command-center/docs/ai-threats.md.txt).