Irregular (formerly Pattern Labs) announces the launch of its Frontier AI Security initiative, positioning itself as the first frontier security lab dedicated to protecting the world as AI systems become increasingly capable and sophisticated.
Irregular has observed artificial intelligence advancing faster than anticipated. The breakthroughs that bring progress in medicine, science, and software also carry risks: systems capable of discovering exploits, evading defenses, or executing autonomous cyber operations. These risks are not hypothetical; they are the natural byproduct of pushing intelligence to its limits.
To address these challenges, Irregular has built a sophisticated research platform to run controlled simulations on frontier AI models. Through these simulations, the company tests both AI models’ potential for misuse in cyber operations and their resilience when they are targeted by attackers.
Already, Irregular’s work has contributed significantly to the field:
-
OpenAI cites Irregular’s evaluations in its system cards for GPT-3, GPT-4-mini, and GPT-5.
-
Irregular is designing and building critical defenses, including a white paper with Anthropic on confidential inference systems.
-
Google DeepMind researchers have cited Irregular in their paper on emerging AI cyberattack capabilities and utilized Irregular’s platform.
-
Irregular co-authored a seminal paper with RAND on securing AI model weights and preventing model theft, helping to set the agenda for global AI security policy and practice together with RAND.
-
Irregular acts as a partner to governmental institutions, such as the UK government, on vetting cyber capabilities in frontier models.
Also Read: CrowdStrike Launches Falcon Fall Release for Agentic Era
The company has secured US $80 million in funding, led by its partners at Sequoia and Redpoint, and is generating millions in annual revenue. Irregular is building practical defensive systems designed to stop threats wherever AI is created or deployed—it emphasizes that this work is not theoretical, but focused on real-world applications.
By uniting applied AI research with nation-state cybersecurity expertise, Irregular is pioneering frontier AI security as a critical new category—becoming the first lab of its kind.
Irregular consists of a small, highly concentrated team. Its members include AI scientists published in top international venues; senior leaders with experience scaling R&D organizations into the hundreds; world debate champions; and leading experts in cyber and national security who have uncovered vulnerabilities in critical infrastructure.
Source: Irregular