diff --git a/packages/security_ai_prompts/changelog.yml b/packages/security_ai_prompts/changelog.yml
index 7ab6969cbce..fef86181214 100644
--- a/packages/security_ai_prompts/changelog.yml
+++ b/packages/security_ai_prompts/changelog.yml
@@ -1,4 +1,9 @@
 # newer versions go on top
+- version: "1.0.14"
+  changes:
+    - description: "Update Entity Highlights AI prompt: add MITRE ATT&CK / Kill Chain mapping, anti-fabrication guardrails, and response size limits"
+      type: enhancement
+      link: https://github.com/elastic/integrations/pull/19999
 - version: "1.0.13"
   changes:
     - description: "Update Entity Highlights AI prompt"
diff --git a/packages/security_ai_prompts/kibana/security_ai_prompt/security_ai_prompts-a14735d2-53d9-4cac-95b5-bc93267eac3f.json b/packages/security_ai_prompts/kibana/security_ai_prompt/security_ai_prompts-a14735d2-53d9-4cac-95b5-bc93267eac3f.json
index 80c67cdc362..e41cd02504a 100644
--- a/packages/security_ai_prompts/kibana/security_ai_prompt/security_ai_prompts-a14735d2-53d9-4cac-95b5-bc93267eac3f.json
+++ b/packages/security_ai_prompts/kibana/security_ai_prompt/security_ai_prompts-a14735d2-53d9-4cac-95b5-bc93267eac3f.json
@@ -3,7 +3,7 @@
     "promptId": "entityDetailsHighlights",
     "promptGroupId": "aiForEntityDetails",
     "prompt": {
-      "default": "Generate structured information for entity so a Security analyst can act. Your response should take all the important elements of the entity into consideration.\n\nGenerate a list of highlight items, each with a title and text. Only include highlights for which information is available in the context.\n  - Risk score: Summarize the entity's risk score and the main factors contributing to it. Don't mention any risk contribution scores.\n  - Criticality: Note the entity's criticality level and its impact on the risk score. Take into account the criticality contribution score inside risk score.\n  - Anomalies: Summarize unusual activities or anomalies detected for the entity and briefly explain why it is significant.\n  - Vulnerabilities: Summarize any significant Vulnerability and briefly explain why it is significant.\n\nAdditionally, provide a list of actionable recommendations for the security analyst if available.\n\n**Guidelines**:\n  - Only include highlight items for which information is available in the context.\n  - Use must use inline code (backticks) for technical values like file paths, process names, arguments, scores, package versions, etc.\n  - **Do not** include any extra explanation, reasoning or text.\n"
+      "default": "Generate structured information for an entity so a Security analyst can act. Your response must take all important elements of the entity context into consideration.\n\nGenerate a list of highlight items, each with a title and text. Only include highlights for which information is available in the context. Keep each highlight to 1 sentence — at most 2, and only when an anomaly needs the extra clause for a MITRE ATT&CK / Kill Chain mapping. Aim to keep the highlights section under 600 characters total.\n  - Risk score: State the score and describe the dominant threat pattern — do not list individual rules or alerts. Only mention a specific rule if it clearly accounts for the majority of the score.\n  - Criticality: State the entity's criticality level and how it affects the overall risk.\n  - Anomalies: Identify the most significant pattern across anomalies rather than listing them. Only if one or more ML job results clearly correspond to a known attack technique, map it to the relevant MITRE ATT&CK tactic (e.g. `Execution`, `Lateral Movement`) or Lockheed Martin Kill Chain phase (e.g. `Exploitation`, `Command & Control`) in the same highlight. If the anomalies are ambiguous or look benign, omit the mapping rather than guessing.\n  - Vulnerabilities: State the most critical vulnerability present and why it matters.\n\nAdditionally, provide up to 3 actionable recommendations for the security analyst, prioritised by urgency. Each must be 1 sentence.\n\n**Guidelines**:\n  - Only include highlight items for which information is available in the context.\n  - Only use values that are explicitly present in the provided context. Do not infer, extrapolate, or fabricate any values, scores, CVEs, job names, or attack-technique mappings that are not present in the context.\n  - You must always use inline code (backticks) for all technical values — criticality levels, risk scores, job names, CVE IDs, CVSS scores, process names, file paths, package versions. Never use single quotes or plain text for these values.\n  - Round all numeric values to 2 decimal places (e.g. `72.00`, `26.62`).\n  - Synthesise — do not list. If multiple signals point to the same pattern, say what the pattern is.\n  - **Do not** include any extra explanation, reasoning or text.\n"
     }
   },
   "id": "security_ai_prompts-a14735d2-53d9-4cac-95b5-bc93267eac3f",
diff --git a/packages/security_ai_prompts/manifest.yml b/packages/security_ai_prompts/manifest.yml
index cbf7a082d70..04484567697 100644
--- a/packages/security_ai_prompts/manifest.yml
+++ b/packages/security_ai_prompts/manifest.yml
@@ -22,4 +22,4 @@ source:
   license: "Elastic-2.0"
 title: "Security AI Prompts"
 type: content
-version: 1.0.13
+version: 1.0.14