Irregular raises $80 million to strengthen security for frontier AI models
Irregular, an AI security company formerly known as Pattern Labs, has raised $80 million in new funding to expand its efforts in safeguarding artificial intelligence systems. The round was led by Sequoia Capital and Redpoint Ventures, with additional participation from Wiz CEO Assaf Rappaport, valuing the company at $450 million, according to a source familiar with the deal.
Securing the Next Phase of AI
The company’s co-founder Dan Lahav believes AI security will soon become one of the defining challenges of the industry.
“Soon, a lot of economic activity is going to come from human-on-AI interaction and AI-on-AI interaction,” Lahav told TechCrunch. “That’s going to break the security stack along multiple points.”
Irregular has already established itself as a major player in AI evaluations. Its work is cited in security assessments of frontier models, including Anthropic’s Claude 3.7 Sonnet and OpenAI’s o3 and o4-mini. Its SOLVE framework, designed to measure a model’s ability to detect vulnerabilities, has become widely used across the industry.
From Existing Risks to Emerging Threats
While Irregular’s initial focus has been on identifying risks in existing models, its new funding will support a broader mission: predicting emergent threats before they reach real-world deployment.
To achieve this, the company has built sophisticated simulated environments where AIs act as both attackers and defenders, stress-testing the resilience of new models.
“When a new model comes out, we can see where the defenses hold up and where they don’t,” explained co-founder Omer Nevo.
Rising Security Concerns Across AI
The push comes at a moment of heightened concern over AI’s dual-use capabilities. Models are growing more adept at discovering software vulnerabilities, a skill that could aid both cyber defenders and malicious actors. Tech leaders have increasingly recognized this risk: OpenAI revamped its internal security measures this summer, partly in response to fears of corporate espionage.
A Moving Target
Irregular’s founders see their role as inseparable from the progress of frontier labs.
“If the goal of the frontier lab is to create increasingly more sophisticated and capable models, our goal is to secure these models,” Lahav said. “But it’s a moving target, so inherently there’s much, much, much more work to do in the future.”
With its fresh $80 million in funding, Irregular is positioning itself at the heart of AI security — aiming to anticipate the vulnerabilities of tomorrow’s most powerful models before they reach the public.
