We send google/gemini-3-flash requests with zeroDataRetention: true, which routes them to Vertex AI.

On that Vertex account:

Is Gemini’s implicit caching disabled? The endpoint list shows implicit caching on the Vertex route, and Google’s docs say ZDR requires caching off. Is the account exempt from Google’s abuse-monitoring prompt logging, through invoiced billing or an approved exception?

Our inputs include audio, sent as a base64 file in Chat Completions, so please confirm both answers cover it.

Thanks!