Detects the first time in 14 days that a Vertex AI caller (user or service account) runs model inference (GenerateContent, StreamGenerateContent, Predict, or RawPredict) from a source autonomous system (ASN). Stolen Google Cloud credentials, including exported service account keys and OAuth tokens, are used for LLMjacking from attacker infrastructure, which shows up as the victim identity calling models from a hosting provider, VPN, or residential proxy network it has never used before. The rule relies on Vertex AI audit logs (aiplatform.googleapis.com) and requires fields client.user.email and source.as.number to be populated by the integration. It detects a new ASN for the same caller within a 14-day history window, verifying that the inference action is one of the supported endpoints and that the log indicates a successful call. The query targets events where event.action matches GenerateContent, StreamGenerateContent, Predict, or RawPredict, the service name is aiplatform.googleapis.com, and the caller identity and ASN are present. This behavior is aligned with attempts to abuse valid cloud accounts to perform AI model inferences from attacker-controlled infrastructure, a pattern associated with LLMjacking. The rule maps to MITRE ATT&CK: Initial Access via Valid Accounts (T1078) with the Cloud Accounts subtechnique (T1078.004) and MITRE ATLAS AML.T0012 (Valid Accounts). It also includes contextual investigation cues and a practical triage/remediation path, such as inspecting source ASN details, validating whether the identity is human or a service account, reviewing user agent history for shifts in tooling, and auditing for IAM changes or key/token activity. Recommendations include disabling or revoking compromised credentials, reviewing IAM bindings that grant aiplatform.endpoints.predict, monitoring Vertex AI usage and billing, and enforcing quotas or budgets to curb abuse. False positives cover legitimate network mobility (e.g., travel or new ISP) and workloads that legitimately move across networks; setup requires Vertex AI audit logs with Data Access enabled and populated client.user.email and source.as.number fields.