Return to Threats

Google Widens Gemini 4 Carbon Access After AI Agents Verify Signal Code

securityweek.com 2026-11-10 AI agent abuse Medium

Executive Summary

Google expanded access to its Carbon platform after DeepMind used Carbon-built agents to verify libsignal security guarantees. Organizations deploying AI agents should assess agent behavior and authorization controls.

What Happened

Google opened Carbon to paying developers and businesses as DeepMind used agents built on it to prove libsignal security guarantees. The post Google Widens Gemini 4 Carbon Access After AI Agents Verify Signal Code appeared first on SecurityWeek .

Affected Products

  • Gemini 4 Carbon
  • Carbon
  • libsignal

Vulnerability / Attack Technique

AI agent verification and code analysis

Technical Mechanism

DeepMind used agents built on Carbon to prove security guarantees for libsignal. The article does not describe the agents’ internal methods, permissions, vulnerabilities, or attack behavior.

AI Security Category

LLM06:2025 Excessive Agency

Part of the AI Agent Excessive Agency topic.

Relevant Frameworks

LLM06:2025 Excessive Agency

RealGround Analysis

The reported use case involves AI agents performing security-related verification, making agent scope and authorization relevant to teams running LLM applications. RealGround analysis: business-logic review and adversarial testing can help ensure agents cannot perform actions beyond their intended verification tasks.

Recommended Controls

  • Define explicit tool, data, and code-access permissions for each AI agent.
  • Require human approval for high-impact code changes or security conclusions.
  • Log agent prompts, tool calls, code modifications, and verification results.
  • Test agent behavior with adversarial scenarios before production use.
  • Periodically audit whether agent actions remain within their documented business purpose.

Related RealGround Service

AI Agent Business Logic Audit · or start with the free AI Agent Security Scanner.

Sources

securityweek.com: https://www.securityweek.com/google-widens-gemini-4-carbon-access-after-ai-agents-verify-signal-code/

Summary and analysis written by RealGround from the source above; last updated 2026-11-11.

Talk to AI CISO