Frontier AI Security

Irregular works with leading AI labs and government institutions to understand the emerging capabilities and security risks of frontier AI systems.

The company builds high-fidelity environments and evaluations that test advanced models and agents against realistic cyber scenarios, measuring capabilities that are difficult to capture through traditional benchmarks.

Its work has been used in frontier model evaluations by organizations including OpenAI, Anthropic, Google DeepMind and Meta.

As AI becomes a more capable tool, target, and actor in cyber operations, Irregular researches the resulting security challenges and develops the environments, methods, and security systems needed to understand and address them.

With one of the largest early funding rounds in the AI security category, Irregular has the capital, talent pipeline, and investor backing to become a defining player in securing the next generation of AI infrastructure.

Irregular, founded in 2023, is backed by Sequoia Capital, Redpoint Ventures, Swish Ventures, and cybersecurity leaders including Assaf Rappaport, CEO of Wiz.

In September 2025, the company announced $80 million in combined Seed and Series A funding.