AI security startup Irregular has secured $80 million in funding to address the critical vulnerabilities inherent in frontier AI models. As large language models (LLMs) grow in complexity, the company aims to provide robust defensive frameworks to protect these systems from emerging cyber threats, positioning itself as a vital player in the race to secure next-generation artificial intelligence.
Simulating the AI Battlefield
The core of Irregular’s approach involves sophisticated network simulations. Co-founder Omer Nevo explains that the platform deploys AI agents to act as both attackers and defenders simultaneously. This adversarial testing allows developers to identify exactly where a model’s defenses hold firm and where they collapse under pressure the moment a new architecture is released.
The Rising Stakes of AI Security
The urgency for such technology has never been higher. As AI systems become increasingly proficient at identifying and exploiting software vulnerabilities, the risks for both corporate and national security have skyrocketed. The industry is responding: OpenAI, for instance, overhauled its internal security protocols this past summer specifically to mitigate the threat of corporate espionage and unauthorized model access.
A Perpetual Arms Race
For the founders of Irregular, this funding round is merely the beginning of a long-term strategy to combat the security challenges posed by rapidly evolving models. The threat landscape is not static; it is a moving target that requires constant adaptation.
Future-Proofing Frontier Tech
“If the goal of the frontier lab is to create increasingly more sophisticated and capable models, our goal is to secure these models,” says co-founder Lahav. Given the exponential pace of AI development, the team acknowledges that the work is far from finished. The company intends to leverage this capital to expand its defensive capabilities, ensuring that as AI grows more intelligent, it does not also become a greater liability for those deploying it.