AI Threat AlertGemini AI Security Vulnerabilities
CVEs affecting Google Gemini, the Gemini API, and Gemini CLI tooling. We track issues where Gemini is the affected component or where Gemini-based integrations are the attack vector.
6
Total CVEs
3
Critical
2
High
Gemini security: where the risk concentrates
Google Gemini exposes risk through three main surfaces. The Gemini CLI is the most attacker-relevant — it executes commands on developer machines and integrates with MCP tools, opening up classic agent attack patterns (prompt-injection-to-RCE chains, untrusted tool argument flow). The Gemini API surface is managed by Google but is consumed by third-party integrations that can leak credentials. Finally, Gemini-powered features in Google Workspace and Search introduce content-injection risks into trusted enterprise contexts.
All tracked Gemini CVEs
| CVE | Severity | Headline / Summary | Published |
|---|---|---|---|
| CVE-2026-30886 | MEDIUM | AI component: IDOR enables unauthorized data access | Mar 23, 2026 |
| CVE-2026-34511 | HIGH | OpenClaw: PKCE verifier leak enables OAuth token theft | Apr 4, 2026 |
| GHSA-wpqr-6v78-jr5g | CRITICAL | Gemini CLI: RCE via malicious workspace in CI/CD | Apr 24, 2026 |
| CVE-2026-12537 | CRITICAL | Gemini CLI: OS command injection enables pre-sandbox RCE | Jun 24, 2026 |
| CVE-2026-13745 | CRITICAL | Gemini CLI: RCE via malicious .env GEMINI_CLI_HOME override | Sep 10, 2026 |
| CVE-2026-117560 | HIGH | ATXP Gemini CLI ext: command injection via MCP tools | Oct 27, 2026 |