Google Vertex AI SDK Flaw Let Attackers Hijack Model Uploads via Bucket Squatting

Palo Alto Networks Unit 42 discovered a predictable-bucket-name vulnerability in the Google Cloud Vertex AI Python SDK that allowed attackers who only knew a victim's project ID to create the expected staging bucket, receive model uploads, replace them with malicious pickle/joblib artifacts, and achieve code execution inside Vertex AI serving containers; Google patched the issue in google-cloud-aiplatform v1.148.0 and recommends setting an explicit staging_bucket and updating the SDK.