Cross-Service Api Key Reuse And Retroactive Permission Expansion
Sources: 1 • Confidence: High • Updated: 2026-03-02 19:32
Key takeaways
- A single Google Cloud API key can be shared across Gemini, Google Maps, and other Google services.
- The corpus recommends that developers audit their API keys for potential cross-service Gemini access exposure.
- Truffle Security identified 2,863 API keys in the November 2025 Common Crawl that could access Gemini, verified via calls to the "/models" listing endpoint.
- A developer can accidentally enable Gemini billing on an existing API key that has already been publicly exposed.
- Enabling the Gemini API within the same Google Cloud project can retroactively expand an existing key's permissions, functioning like privilege escalation over time.
Sections
Cross-Service Api Key Reuse And Retroactive Permission Expansion
- A single Google Cloud API key can be shared across Gemini, Google Maps, and other Google services.
- A developer can accidentally enable Gemini billing on an existing API key that has already been publicly exposed.
- Enabling the Gemini API within the same Google Cloud project can retroactively expand an existing key's permissions, functioning like privilege escalation over time.
Remediation And Operational Monitoring
- The corpus recommends that developers audit their API keys for potential cross-service Gemini access exposure.
- Google is working to revoke affected API keys.
Internet-Scale Detectability Of Exposed Gemini-Capable Keys
- Truffle Security identified 2,863 API keys in the November 2025 Common Crawl that could access Gemini, verified via calls to the "/models" listing endpoint.
Watchlist
- The corpus recommends that developers audit their API keys for potential cross-service Gemini access exposure.
Unknowns
- What exact Google Cloud project settings and API enablement steps cause an existing API key to gain Gemini access (and/or billing exposure) retroactively?
- What is the practical scope of what an exposed Gemini-capable key can do beyond listing models via the "/models" endpoint (e.g., invoking billable operations or accessing data)?
- How many of the identified exposed keys remain valid over time, and how quickly does Google invalidate them once identified?
- What safeguards (if any) prevent or limit cross-service key reuse risk (e.g., least-privilege controls, service-specific restrictions), and how effective are they in this scenario?
- Is there any direct decision-readthrough (operator, product, or investor) beyond generic key auditing and rotation guidance?