Detects OpenAI ChatGPT Enterprise assistant messages that recognize likely exposed secrets in chat and advise rotating, revoking, regenerating, invalidating, or deactivating credentials (API keys, access keys, passwords, tokens, and similar). The rule analyzes assistant messages collected by the OpenAI ChatGPT Enterprise integration and flags instances where credential-related terms (e.g., API key, secret, password, token) appear in combination with explicit exposure cues and rotation/advice language (rotate, revoke, regenerate, invalidate, deactivate). A live secret exposure in conversation triggers a high-severity alert and is mapped to MITRE ATT&CK Unsecured Credentials (T1552) and MITRE ATLAS AML.T0055/AML.T0057 signals for credential access and data leakage. It supports triage by correlating with prior user messages and other in-conversation credential alerts, and guides remediation to rotate/revoke the exposed credential and restrict subsequent access to alert artifacts. This rule requires the OpenAI ChatGPT Enterprise integration to collect conversation message logs into Elasticsearch and uses an ES|QL query to scan assistant messages for the specified patterns.