Live / Google Gemini / Incident
Google Gemini: Vertex AI Gemini API customers experienced increased error rates when accessing the global endpoint.
Google Gemini · gemini.google.comSource: Vertex AI Gemini APIResolved2026-03-09 05:25 UTC
Public — what happened
Summary
Google Gemini said customers of its Vertex AI Gemini API saw elevated error rates on February 27, 2026, from 04:37 to 06:35 US/Pacific, a span of 1 hour and 58 minutes, affecting the global endpoint and US regions across all Gemini model versions. Related services that depend on Gemini, including Dialogflow CX, Agent Assist and Google Cloud Support, were also affected, with customers mostly receiving 429 or 503 errors. Google attributed the problem to a configuration change made to a safety filtering service used by all Gemini models, and said it reversed that change and added capacity to bring service back. The company said it plans to add required validation steps to its rollout process and to strengthen alerting on critical dependencies.
Timeline
- Started — 2026-02-27 12:37 UTC
- Monitoring — 2026-02-27 16:12 UTC
- Resolved — 2026-02-27 14:35 UTC · 1h 58m after it started
With SquidGrid — where it starts on the route
Independent path health
The status page is the symptom. SquidGrid measures the paths that reach Google Gemini against their own baselines — whether the routes carrying your traffic degraded, at which handoff, and who owns that stretch.
🔒 Sign in to seeCustomers get the live Provider Assessment Report for Google Gemini: its full incident track record, what tends to break, and recent public reports of problems.
Included with any paid plan.
Open the Google Gemini report →
More from Google Gemini