diff --git a/actions/setup/js/parse_codex_log.cjs b/actions/setup/js/parse_codex_log.cjs
index 23a3452978..686cc618b9 100644
--- a/actions/setup/js/parse_codex_log.cjs
+++ b/actions/setup/js/parse_codex_log.cjs
@@ -122,6 +122,42 @@ function extractMCPInitialization(lines) {
   };
 }
 
+/**
+ * Extract error messages from Codex logs (e.g., model access blocked, cyber_policy_violation)
+ * @param {string[]} lines - Array of log lines
+ * @returns {{hasErrors: boolean, messages: string[], reconnectCount: number, maxReconnects: number}} Error info
+ */
+function extractCodexErrorMessages(lines) {
+  const messages = new Set();
+  let reconnectCount = 0;
+  let maxReconnects = 0;
+
+  for (const line of lines) {
+    // Match: ERROR: <message> (final error after all retries exhausted)
+    const errorMatch = line.match(/^ERROR:\s*(.+)$/);
+    if (errorMatch) {
+      messages.add(errorMatch[1].trim());
+    }
+
+    // Match: Reconnecting... N/M (error message) - reconnect attempts with error details
+    const reconnectMatch = line.match(/^Reconnecting\.\.\.\s+(\d+)\/(\d+)\s*\((.+)\)$/);
+    if (reconnectMatch) {
+      const attempt = parseInt(reconnectMatch[1]);
+      const total = parseInt(reconnectMatch[2]);
+      if (attempt > reconnectCount) reconnectCount = attempt;
+      if (total > maxReconnects) maxReconnects = total;
+      messages.add(reconnectMatch[3].trim());
+    }
+  }
+
+  return {
+    hasErrors: messages.size > 0,
+    messages: Array.from(messages),
+    reconnectCount,
+    maxReconnects,
+  };
+}
+
 /**
  * Convert parsed Codex data to logEntries format for plain text rendering
  * @param {Array<{type: string, content?: string, toolName?: string, params?: string, response?: string, statusIcon?: string}>} parsedData - Parsed Codex log data
@@ -255,6 +291,18 @@ function parseCodexLog(logContent) {
     markdown += mcpInfo.markdown;
   }
 
+  // Extract error messages (e.g., model access blocked, cyber_policy_violation)
+  const errorInfo = extractCodexErrorMessages(lines);
+  if (errorInfo.hasErrors) {
+    markdown += "## ⚠️ Errors\n\n";
+    for (const message of errorInfo.messages) {
+      markdown += `> ${message}\n\n`;
+    }
+    if (errorInfo.reconnectCount > 0) {
+      markdown += `> Reconnect attempts: ${errorInfo.reconnectCount}/${errorInfo.maxReconnects}\n\n`;
+    }
+  }
+
   markdown += "## 🤖 Reasoning\n\n";
 
   // Second pass: process full conversation flow with interleaved reasoning and tools
@@ -662,5 +710,6 @@ if (typeof module !== "undefined" && module.exports) {
     formatCodexToolCall,
     formatCodexBashCall,
     extractMCPInitialization,
+    extractCodexErrorMessages,
   };
 }
diff --git a/actions/setup/js/parse_codex_log.test.cjs b/actions/setup/js/parse_codex_log.test.cjs
index 399f71ad8a..481b82802a 100644
--- a/actions/setup/js/parse_codex_log.test.cjs
+++ b/actions/setup/js/parse_codex_log.test.cjs
@@ -542,4 +542,120 @@ I will analyze the code`;
       expect(result.markdown).toContain("## 🤖 Reasoning");
     });
   });
+
+  describe("extractCodexErrorMessages function", () => {
+    let extractCodexErrorMessages;
+
+    beforeEach(async () => {
+      const module = await import("./parse_codex_log.cjs");
+      extractCodexErrorMessages = module.extractCodexErrorMessages;
+    });
+
+    it("should extract ERROR: lines", () => {
+      const lines = [
+        "thinking",
+        "Some thinking content here",
+        "ERROR: stream disconnected before completion: This user's access to gpt-5.3-codex has been temporarily limited for potentially suspicious activity related to cybersecurity. Learn more about our safety mitigations: https://platform.openai.com/docs/guides/safety-checks/cybersecurity",
+      ];
+
+      const result = extractCodexErrorMessages(lines);
+
+      expect(result.hasErrors).toBe(true);
+      expect(result.messages).toHaveLength(1);
+      expect(result.messages[0]).toContain("stream disconnected before completion");
+      expect(result.messages[0]).toContain("cyber");
+    });
+
+    it("should extract Reconnecting messages with error details", () => {
+      const lines = [
+        "Reconnecting... 1/5 (stream disconnected before completion: This user's access to gpt-5.3-codex has been temporarily limited for potentially suspicious activity related to cybersecurity. Learn more about our safety mitigations: https://platform.openai.com/docs/guides/safety-checks/cybersecurity)",
+        "Reconnecting... 2/5 (stream disconnected before completion: This user's access to gpt-5.3-codex has been temporarily limited for potentially suspicious activity related to cybersecurity. Learn more about our safety mitigations: https://platform.openai.com/docs/guides/safety-checks/cybersecurity)",
+      ];
+
+      const result = extractCodexErrorMessages(lines);
+
+      expect(result.hasErrors).toBe(true);
+      expect(result.messages).toHaveLength(1); // De-duplicated
+      expect(result.reconnectCount).toBe(2);
+      expect(result.maxReconnects).toBe(5);
+    });
+
+    it("should de-duplicate identical error messages from retries", () => {
+      const errorMsg = "stream disconnected before completion: This user's access to gpt-5.3-codex has been temporarily limited";
+      const lines = [`Reconnecting... 1/5 (${errorMsg})`, `Reconnecting... 2/5 (${errorMsg})`, `Reconnecting... 3/5 (${errorMsg})`, `ERROR: ${errorMsg}`];
+
+      const result = extractCodexErrorMessages(lines);
+
+      expect(result.hasErrors).toBe(true);
+      expect(result.messages).toHaveLength(1);
+      expect(result.reconnectCount).toBe(3);
+    });
+
+    it("should return no errors for normal log lines", () => {
+      const lines = ["thinking", "I will list the open pull requests", "tool github.list_pull_requests({})"];
+
+      const result = extractCodexErrorMessages(lines);
+
+      expect(result.hasErrors).toBe(false);
+      expect(result.messages).toHaveLength(0);
+      expect(result.reconnectCount).toBe(0);
+      expect(result.maxReconnects).toBe(0);
+    });
+
+    it("should handle empty lines array", () => {
+      const result = extractCodexErrorMessages([]);
+
+      expect(result.hasErrors).toBe(false);
+      expect(result.messages).toHaveLength(0);
+    });
+  });
+
+  describe("parseCodexLog with model access errors", () => {
+    it("should include Errors section when ERROR: line is present", () => {
+      const logContent = `2026-02-27T14:06:41.886993Z  INFO session_loop: codex_core::codex: Turn error: stream disconnected before completion: This user's access to gpt-5.3-codex has been temporarily limited for potentially suspicious activity related to cybersecurity. Learn more about our safety mitigations: https://platform.openai.com/docs/guides/safety-checks/cybersecurity
+ERROR: stream disconnected before completion: This user's access to gpt-5.3-codex has been temporarily limited for potentially suspicious activity related to cybersecurity. Learn more about our safety mitigations: https://platform.openai.com/docs/guides/safety-checks/cybersecurity`;
+
+      const result = parseCodexLog(logContent);
+
+      expect(result.markdown).toContain("## ⚠️ Errors");
+      expect(result.markdown).toContain("stream disconnected before completion");
+      expect(result.markdown).toContain("cybersecurity");
+    });
+
+    it("should include Errors section with reconnect count when retries occurred", () => {
+      const logContent = `Reconnecting... 1/5 (stream disconnected before completion: This user's access to gpt-5.3-codex has been temporarily limited for potentially suspicious activity related to cybersecurity)
+Reconnecting... 2/5 (stream disconnected before completion: This user's access to gpt-5.3-codex has been temporarily limited for potentially suspicious activity related to cybersecurity)
+Reconnecting... 3/5 (stream disconnected before completion: This user's access to gpt-5.3-codex has been temporarily limited for potentially suspicious activity related to cybersecurity)
+ERROR: stream disconnected before completion: This user's access to gpt-5.3-codex has been temporarily limited for potentially suspicious activity related to cybersecurity`;
+
+      const result = parseCodexLog(logContent);
+
+      expect(result.markdown).toContain("## ⚠️ Errors");
+      expect(result.markdown).toContain("Reconnect attempts: 3/5");
+    });
+
+    it("should not include Errors section for normal logs", () => {
+      const logContent = `thinking
+I will list the open pull requests
+tool github.list_pull_requests({"state":"open"})
+github.list_pull_requests(...) success in 123ms:
+{"items": []}`;
+
+      const result = parseCodexLog(logContent);
+
+      expect(result.markdown).not.toContain("## ⚠️ Errors");
+    });
+
+    it("should place Errors section before Reasoning section", () => {
+      const logContent = `ERROR: This user's access to gpt-5.3-codex has been temporarily limited`;
+
+      const result = parseCodexLog(logContent);
+
+      const errorsIndex = result.markdown.indexOf("## ⚠️ Errors");
+      const reasoningIndex = result.markdown.indexOf("## 🤖 Reasoning");
+      expect(errorsIndex).toBeGreaterThan(-1);
+      expect(reasoningIndex).toBeGreaterThan(-1);
+      expect(errorsIndex).toBeLessThan(reasoningIndex);
+    });
+  });
 });