Google Broadens Availability of Gemini 4 Carbon Following AI Verification Breakthroughs
Google has announced the commercial expansion of its advanced model, Gemini 4 Carbon, making the technology available to paying developers and business users. The rollout extends access through the Gemini API and Gemini for business channels, while Google AI Ultra subscribers receive immediate access. Introductory pricing structures have been established through the end of the month before standard rates take effect.
Expanding Enterprise and Developer Access
Previously restricted to specialized defenders participating in Google’s Fairwind Program, Gemini 4 Carbon is now accessible to a broader market, though specific cyber capabilities still require Fairwind membership. Industry evaluations indicate strong performance metrics for the technology. For instance, security firm Wiz utilized Carbon for cyber tools to identify zero-day vulnerabilities across major software-as-a-service providers within a remarkably short timeframe. Similarly, internal teams leveraged the model to confirm the exploitability of open-source project flaws, enabling maintainers to issue critical patches swiftly.
Formal Verification of Complex Code
In a related development, Google DeepMind introduced AlphaProof Code, an initiative pairing human expertise with automated agents to execute formal software verification. By combining human oversight with models such as Carbon and Claude Opus 5.5, researchers tackled the reference implementation of the Signal protocol, known as libsignal. Automated agents successfully generated millions of lines of mathematical proofs, which were rigorously checked for validity.
- Proved critical guarantees including forward secrecy and authentication for compiled machine code
- Reduced a verification timeline that traditionally spans years down to a matter of weeks
- Demonstrated that artificial intelligence can make formal verification practical for enterprise systems
Signal leadership noted that rigorously verifying implementation details provides heightened confidence in core privacy guarantees. Google has made the resulting specifications and proofs publicly available, marking a significant step toward practical, large-scale code verification.