Google expands access to Gemini 4 Carbon following AI agent verification

Google has expanded access to its Carbon technology, making it available to paying developers and businesses. According to SecurityWeek, this move follows successful testing by DeepMind, which utilized AI agents built on the platform to verify the security guarantees of libsignal code.

The expansion of Carbon access follows the demonstrated ability of DeepMind's AI agents to perform technical verification tasks. The agents were employed to prove the security properties of the libsignal codebase, a process that informs the broader availability of the tool to commercial users.

The scope of the update involves making Carbon available to a wider range of users within the developer and business sectors. This transition from internal DeepMind testing to broader commercial availability marks a shift in how the technology is deployed for security verification.

The context regarding the application of these AI agents within the development and security sectors is provided by SecurityWeek. The reporting focuses on the intersection of large language model capabilities and the verification of cryptographic or security-focused codebases.