Stolen Gemini API Key Generates $82,000 Bill in 48 Hours, Exposing Cloud Security Flaw

Gemini API Key Theft Threatens Small Business With Bankruptcy

A three-person development team in Mexico is confronting financial ruin after a stolen Gemini API key generated $82,314.44 in unauthorized charges between February 11 and February 12, 20261

1

. The company, which typically spends $180 per month on Google Cloud services, experienced a 455× usage spike in just 48 hours4

4

. Developer RatonVaquero shared their predicament on Reddit, stating the firm faces bankruptcy if Google enforces even a third of these charges. The unauthorized API usage primarily involved Gemini 3 Pro Image and Gemini 3 Pro Text services, with thieves exploiting the compromised credentials to generate massive volumes of AI content on the victim's account2

2

.

Google Invokes Shared Responsibility Model Amid Cloud Billing Crisis

Google representatives have cited the company's Shared Responsibility Model when addressing the $82,000 in charges, indicating the developer may remain liable for the unauthorized API usage1

1

. Under this framework, Google Cloud secures its platform while users must implement authentication systems, access policies, and network security to protect their credentials2

2

. The affected developers have since deleted the compromised key, disabled Gemini APIs, rotated credentials, enabled two-factor authentication, locked down IAM, and filed a cybercrime report with the FBI1

1

. Despite these remediation efforts, initial support interactions suggest Google may not offer payment adjustments, leaving the small business in financial jeopardy.

API Key Vulnerability Exposes 2,863 Live Google API Keys

Truffle Security researchers uncovered a systemic API key vulnerability affecting thousands of organizations. Their investigation revealed 2,863 live public Google API keys with inadvertent Gemini access2

2

. The security flaw stems from how Google Cloud API keys, originally designed as project identifiers for billing purposes, now authenticate to Gemini endpoints after the generative AI service is enabled3

3

. These keys, identifiable by the "AIza" prefix, were embedded in client-side code for services like Google Maps and Firebase based on Google's earlier guidance that API keys were not secrets3

3

. Truffle Security researcher Joe Leon explained that attackers with valid keys can access uploaded files, cached data via /files and /cachedContents API endpoints, and charge LLM usage to victim accounts3

3

.

Legacy Credentials Create Expanding Attack Surface

The vulnerability highlights how legacy credentials gain sensitive privileges without developer awareness. When users enable the Gemini API on a Google Cloud project through the Generative Language API, existing API keys in that project automatically gain access to Gemini endpoints without warning. Additionally, creating new API keys in Google Cloud defaults to "Unrestricted" status, making them applicable for every enabled API in the project, including Gemini3

3

. Quokka's separate research scanning 250,000 Android apps found over 35,000 unique Google API keys embedded in mobile applications3

3

. This transformation of harmless billing tokens into live authentication credentials represents what security experts call a dynamic risk scenario where generative AI security concerns intersect with cloud infrastructure vulnerabilities5

5

.

Developers Demand Catastrophic Usage Anomalies Protections

Affected developers are calling for basic spending guardrails against catastrophic usage anomalies, arguing that cloud providers should implement automatic safeguards when billing patterns deviate dramatically from established norms1

1

. The Mexican development team noted that personal Gemini customers have usage caps preventing overspending, while business users can set quotas limiting requests per day or minute, and Google Cloud Vertex AI users can configure budget alerts1

1

. However, these protections require proactive configuration and don't include mandatory spending caps that would prevent unauthorized API usage from generating ruinous bills. The developers argue that features like temporary service freezing pending review and mandatory verification for extreme billing spikes should be standard protections.

Google Responds With Proactive Detection Measures

After Truffle Security presented their findings through Google's Vulnerability Disclosure Project, the company initially classified the report as "intended behavior" before upgrading it to "Bug" status in December 2025 when researchers provided examples from Google's own infrastructure2

2

. A Google spokesperson stated the company has "implemented proactive measures to detect and block leaked API keys that attempt to access the Gemini API"3

3

. However, as of February 2026, Truffle Security reported not seeing a concrete outcome from the root-cause fix Google was developing2

2

. Organizations using Google Cloud should verify if AI-related APIs are enabled in their projects and rotate keys that may be publicly accessible, prioritizing oldest keys first since they were most likely deployed under previous guidance treating API keys as safe to share3

3

. Truffle Security offers TruffleHog, an open-source secrets scanning tool, to help identify exposed credentials across code repositories and web assets2

2

.