Google Cloud Project ID: clarity-339fc
App: Clarity, a policy-library app (Next.js) built in AI Studio.
Error on every Publish attempt (step 1): “Failed to generate API key, The request is suspicious. Please try again.” The AI Studio deployment log also shows: “Deployment failed with the following errors. Fix the applet code based on these logs: Suspicious activity detected. Request blocked for project clarity-339fc.”
On a later attempt, after retrying past that, it progressed further and failed with: “Failed to create Cloud Run service: Precondition check failed.”
What I’ve already verified: Billing is active and paid on this project. Only 2-3 total GCP projects exist on the account, nowhere near any project-count limit. Cloud Run services and domain mappings show nothing but one unrelated, pre-existing Cloud Function in the console — no orphaned/stale resources visible there. No organization or org policy applies to this project. Cloud Build history shows zero triggered builds, so the block appears to happen before any real build/deploy starts. Re: the unrestricted-API-key announcement pinned above — checked Credentials, every key on this project is already restricted to the Gemini API (including the ais-gemini-key-* keys AI Studio itself generates on each Publish attempt), so that isn’t the cause here.
New today (Sep 15): I pulled Cloud Logging audit entries for this project (filter: protoPayload.methodName=~"CreateService|CreateDomainMapping" severity>=ERROR). In one 10-second window, Publish auto-retried 5 different custom-URL slugs, and every one failed, split across two distinct sub-errors. clarity.ai.studio, clarity-1.ai.studio, and clarity-2.ai.studio each returned “Domain mapping already exists for [slug].ai.studio. Please use a different custom url.” clarity-3141.ai.studio (region us-west1) and get-clarity.ai.studio (region us-east1) each returned “Resource ‘’ of kind ‘ROUTE’ in region ‘[region]’ … does not exist.” There are no CreateService entries at all in that window — it never gets past domain-mapping creation.
That “Resource ‘’ of kind ‘ROUTE’ … does not exist” error is identical, verbatim, to what’s reported (still unresolved after 15 days) in Failed to publish app, which also points to an internal AI Studio/GCP state mismatch rather than anything project-specific. Also flagging my earlier report on this same project in the existing “Publish wizard stuck in infinite loop” thread in case it helps connect the two.
Happy to record the publish attempt on video too — let me know exactly what’s most useful to capture (the initial “suspicious activity” toast, the retry/“Fixing errors” loop, or the later Cloud Run precondition failure) and I’ll get it over.
Update (later same day, Sep 15): Found and fixed something real. This project had been switched to Google’s Cloud Prepay billing model for the Gemini API and sat at a $0.00 credit balance, with the Billing page showing “A credit balance above $0 is required to resume service” and the project’s Status flagged “Prepay required.” Added $5 in prepay credit, balance is now confirmed positive, and that status flag is gone.
With billing fixed, retried Publish using a brand-new custom URL never typed before (use-clarity), completely different from all five prior slugs above. Got through project/billing confirmation and the Gemini API spend-cap step cleanly, then hit the exact same “Failed to create Cloud Run service: Precondition check failed” toast again.
Pulled Cloud Logging again right after (same filter as above, last two days). use-clarity.ai.studio’s CreateDomainMapping call failed with the identical “Resource ‘’ of kind ‘ROUTE’ in region ‘us-east1’ in project ‘clarity-339fc’ does not exist” error as the other slugs. Publish also auto-tried a burst of other slugs I never typed in the same window, all failing the same two ways (already-exists or missing ROUTE). Across the entire log history for this project, there is still not one single CreateService entry — every attempt dies at CreateDomainMapping.
So this rules out both billing and slug reuse as the cause. Whatever is broken is in AI Studio’s own domain-mapping state for this project, independent of funding status and independent of which custom URL is chosen.