AWS guardduty: Correlate AI Protection findings into IAM attack sequences

Service: guardduty · 2026-09-27 · Documentation medium

File: guardduty/latest/ug/guardduty-attack-sequence-finding-types.md · Type: iam

Summary

Adds documentation that enabling AI Protection lets GuardDuty include AI Protection finding types as correlation signals in AttackSequence:IAM/CompromisedCredentials, associating them with an IAM identity and correlating with other suspicious actions.

Security assessment

Documents threat-detection correlation behavior for compromised IAM credentials, a security capability. It is guidance about detection coverage, not a fix for a specific vulnerability.

Evidence

+When you enable [AI Protection](./ai-protection.html), GuardDuty can also include [AI Protection finding types](./findings-ai-protection.html) among the signals that it correlates for this attack sequence. These findings are associated with an IAM identity. As a result, GuardDuty can surface them as signals in this attack sequence and correlate them with other suspicious actions taken with the same potentially compromised credentials.

Diff

diff --git a/guardduty/latest/ug/guardduty-attack-sequence-finding-types.md b/guardduty/latest/ug/guardduty-attack-sequence-finding-types.md
index 266698f1d..0b8d47246 100644
--- a//guardduty/latest/ug/guardduty-attack-sequence-finding-types.md
+++ b//guardduty/latest/ug/guardduty-attack-sequence-finding-types.md
@@ -147,0 +148,2 @@ GuardDuty uses its proprietary correlation algorithms to observe and identify th
+When you enable [AI Protection](./ai-protection.html), GuardDuty can also include [AI Protection finding types](./findings-ai-protection.html) among the signals that it correlates for this attack sequence. These findings are associated with an IAM identity. As a result, GuardDuty can surface them as signals in this attack sequence and correlate them with other suspicious actions taken with the same potentially compromised credentials. For example, GuardDuty can correlate other signals with an anomalous Amazon Bedrock or Amazon SageMaker AI model invocation made with those credentials, and include it in this attack sequence. On their own, AI Protection findings do not generate an attack sequence. For more information, see [AI Protection](./ai-protection.html).
+