Gemini for Google Cloud API seems to run on its own ?

[Rifad]

Hello everyone,

I’m seeing unexpectedly high usage for “Gemini for Google Cloud API” in my project, even though I haven’t deployed any services or made any explicit Generative AI calls in my code. Here’s the situation:

1. No Active Deployments or Scripts

   • I have not written or deployed any Vertex AI / Gemini-based code (no notebooks, no webhooks, no scheduled jobs).

   • I haven’t clicked any “Try this API” buttons in the documentation or run any REST calls myself.

2. API & Services Dashboard

   • When I navigate to APIs & Services → Dashboard, I see that “Gemini for Google Cloud API” (service name cloudaicompanion.googleapis.com) shows very high request counts.

   • Under “Metrics → Credentials,” most of that traffic is listed as Unspecified, which suggests it’s coming from the Cloud Console rather than a named service account or API key.
     

     

3. Cost Impact

   • My billing page shows large token‐count SKUs (e.g. “Generate content input token count Gemini 2.5 Pro Experimental long input text”) with thousands of input/output tokens, resulting in a hefty charge even though I didn’t consciously run any inference.

4. What I’ve Checked So Far

   • I verified that I never manually ran a projects.locations.models.predict RPC.

   • I confirmed there are no Cloud Functions, Compute Engine VMs, or Cloud Run services in this project making AI calls.

   • In IAM & Admin → Audit Logs, “Data Read” was initially disabled for both cloudaicompanion.googleapis.com and aiplatform.googleapis.com, so I wasn’t seeing any “per‐request” logs. (After enabling it, I still don’t see any Vertex AI audit entries related to these charges.)

---

Questions:

1. Why is the Console’s “Gemini for Google Cloud” assistant generating usage when I haven’t clicked any buttons or run any code myself?

2. How can I identify exactly which console action is triggering these token‐billing SKUs?

3. What steps can I take to prevent the Cloud Console from automatically calling Gemini on my behalf (so that I don’t accrue charges without realizing it)?

Any guidance on why “Gemini for Google Cloud API” appears under heavy usage even when I thought I wasn’t using it and how to lock it down would be greatly appreciated. Thank you!

[LeoK]

Hello @Rifad

I’d start by checking the Log Explorer for those specific dates to understand what happened.

My first guess would’ve been someone using Gemini Prompt via Vertex AI but since you’re saying nothing was written into Vertex, the logs should help confirm that.

What really caught my eye is the cloudaicompanion.googleapis.com entry. Check with your team if someone has set up Gemini Code Assist in their IDE. That could explain the activity.