irregular.comCrawled Oct 31, 2026
Irregular (formerly Pattern Labs) is the first frontier security lab with the mission of protecting the world in the time of increasingly capable and sophisticated AI systems.
The company presents a scoring framework to evaluate the complexity of discovering vulnerabilities and creating exploits in end-to-end challenges.
This blog post describes the company’s Evaluation Platform, which tests AI risks through empirical methods and supports top frontier labs in measuring associated threats.
In a new study, agents autonomously leaked credentials and disabled security tools when given open‑ended goals.
Multi-agent AI systems can cooperate to exfiltrate data and bypass access controls, Irregular researchers demonstrated.
Research from Irregular surfaces new attack surfaces that emerge when large and small language models interact inside production systems.
A growing share of frontier chatbots fail to follow operator constraints under adversarial pressure, according to Irregular's latest study.
Yoni Rozenshein’s 2024 BlueHat IL talk outlines a method for assessing AI-driven cyber threats, including a capture-the-flag challenge with large language models, and discusses stakeholders like governments and frontier AI labs.
Irregular is working with NVIDIA on the Open Agent Safety Platform, an open software platform and reference system design for securing AI agents from testing through deployment.
Irregular has been named a winner on Fast Company's 2026 World Changing Ideas Awards, honoring creative solutions to the world's most pressing problems.
Irregular has been named to Redpoint's InfraRed 100, which spotlights 100 companies shaping cloud infrastructure, with Irregular featured among the leaders in security.
Irregular is the first frontier security lab with the mission of protecting the world in the time of increasingly capable and sophisticated AI systems.