{"data":{"id":"6c4653cc-0c44-45f5-9fe2-706baca70759","title":"Help defend against memory poisoning in Amazon Bedrock AgentCore with multi-layer write validation","summary":"Amazon Bedrock AgentCore Memory gives AI agents persistent short-term and long-term memory, and memory content is automatically retrieved into future conversations. The post shows how to compose AgentCore's existing controls (IAM condition keys, namespace-variable validation, consolidation prompts and the ApplyGuardrail API) with an embedding-based semantic-drift signal into a write-time validation layer against memory poisoning, which it describes as a form of indirect prompt injection.","solution":"The post's approach is to compose AgentCore's built-in controls into a defense-in-depth pattern on the memory write path, adding an embedding-based semantic-drift signal and a single ALLOW, QUARANTINE or DENY decision with quarantine isolation and alerting, without changes to agent code.","labels":["security","industry"],"sourceUrl":"https://aws.amazon.com/blogs/security/help-defend-against-memory-poisoning-in-amazon-bedrock-agentcore-with-multi-layer-write-validation/","publishedAt":"2026-11-09T19:05:52.000Z","cveId":null,"cweIds":null,"cvssScore":null,"cvssSeverity":null,"severity":"low","attackType":["prompt_injection","rag_poisoning"],"issueType":"news","affectedPackages":null,"affectedPackageNames":null,"affectedPackageRefs":null,"affectedVendors":["Amazon"],"affectedVendorsRaw":["Amazon Bedrock AgentCore","Amazon Bedrock Guardrails","AWS Identity and Access Management"],"classifierModel":"claude-haiku-5-5","classifierPromptVersion":"v4","summaryPromptVersion":"v2","headline":null,"headlinePromptVersion":null,"cvssVector":null,"attackVector":null,"attackComplexity":null,"privilegesRequired":null,"userInteraction":null,"exploitMaturity":null,"epssScore":null,"epssCheckedAt":null,"kevDateAdded":null,"advisoryAliases":null,"affectedPackagesSource":null,"affectedPackagesCheckedAt":null,"patchAvailable":null,"disclosureDate":"2026-11-09T19:05:52.000Z","fixReleasedAt":null,"daysToFix":null,"fixCheckedAt":null,"capecIds":null,"crossRefCount":0,"attackSophistication":"moderate","impactType":["integrity","safety"],"aiComponentTargeted":"agent","llmSpecific":false,"classifierConfidence":0.85,"researchCategory":null,"atlasIds":["AML.T0051","AML.T0070"]}}