Thousands of Google Cloud API Keys Exposed to Gemini Abuse and Massive Billing Risks

Thousands of Public Google Cloud API Keys Exposed with Gemini Access After API Enablement

Truffle Security discovered nearly 3,000 Google API keys embedded in client-side code that can now be used to authenticate to sensitive Gemini endpoints. This exposure allows attackers to access uploaded files, cached data, and charge LLM usage to victim accounts.

Why This Matters

The technical reality of “Unrestricted” API keys contradicts the ideal model of granular security. When organizations enable the Gemini API on an existing project, legacy billing identifiers used for benign front-end services like Google Maps automatically gain access to high-cost LLM endpoints. This retroactive over-permissioning creates a significant financial and data risk, as evidenced by a reported incident where a stolen key resulted in $82,314.44 in charges within 48 hours, compared to a standard $180 monthly spend.

Key Insights

• 2,863 live ‘AIza’ keys were found publicly accessible in client-side JavaScript, including on a Google-associated website (Truffle Security, 2026).
• Quokka’s scan of 250,000 Android apps identified over 35,000 unique Google API keys embedded in mobile code (Quokka, 2026).
• Default ‘Unrestricted’ status on new Google Cloud API keys makes them applicable for every enabled API in a project, including Gemini.
• Compromised keys allow unauthorized access to sensitive data stored in /files and /cachedContents endpoints.
• Retroactive permissioning occurs when the Generative Language API is enabled, granting existing keys access to Gemini without administrative notice.

Practical Applications

• Use Case: Rotating oldest API keys first as they are the most likely to have been deployed publicly under legacy guidance before gaining retroactive AI privileges.
• Pitfall: Enabling the Generative Language API on projects with existing public keys, which grants surreptitious Gemini access without warning.
• Use Case: Implementing continuous security testing and behavioral profiling to identify anomalies in API data access and quota consumption.
• Pitfall: Relying on API keys as simple billing identifiers for client-side services without applying strict API restrictions in the Google Cloud Console.

References:

• http://thehackernews.com/2026/02/thousands-of-public-google-cloud-api.html