Ctrl+G builds AI security systems that teach frontier models to defend against autonomous cyberattacks. The company works at the model level to make AI-generated code secure by design, and partners with agent companies to design defensive workflows, evaluation harnesses, and custom tooling. Their end goal is an autonomous defender that reviews code, audits infrastructure, and hardens cloud environments in real time.
Funding
Funding not disclosed
Founders
Product
Problem
Autonomous AI agents are creating entirely new categories of cyber offense, probing, adapting, and spreading without human direction. Attackers can now craft exploits with intelligence and build attack chains no human ever designed. Meanwhile, AI-generated code is flooding production faster than any team can review, with insecure patterns repeated at scale across millions of repositories, expanding the attack surface with every commit.
Solution
Ctrl+G works at the model level to ensure AI output is secure by design, transforming capable models into security operators. The company collaborates with agent companies to design workflows around models, including custom tooling, evaluation harnesses, and agentic pipelines. Their approach combines CVE reproduction with adversarial training in CTF environments to create foundations for autonomous defense systems. The end state is an autonomous defender that reviews code as it's written, audits infrastructure configs before deployment, flags misconfigurations in CI/CD pipelines, and hardens cloud environments in real time across the entire stack.
Target Audience
Primary customers are AI agent companies and enterprises deploying AI-generated code at scale that need autonomous security defense integrated at the model level.
Features
- Model-level security training that embeds defensive capabilities directly into AI systems
- CVE reproduction and adversarial training in CTF environments to build practical defense skills
- Custom tooling and evaluation harnesses for agent companies to transform capable models into security operators
- Autonomous defense across source code, Kubernetes manifests, IAM policies, network rules, and runtime behavior
- Real-time code review, infrastructure auditing, and CI/CD misconfiguration flagging
- Continuous cloud environment hardening without human intervention