Irregular

Irregular is a frontier AI security lab based in Tel Aviv, Israel, that conducts red teaming, safety evaluations, and misuse testing on advanced artificial intelligence models. The company has secured approximately $80 million in funding and is projected to reach a valuation of around $450 million, generating millions in revenue through contracts with clients such as OpenAI, Anthropic, and Google DeepMind. It specializes in running controlled simulations to expose AI models to realistic threat scenarios, including cyberattacks and autonomous offensive operations, identifying vulnerabilities prior to public deployment. Lead investors include Sequoia Capital, Assaf Rappaport (Wiz founder), and Ofir Ehrlich (Eon founder). Irregular was founded in 2023 by Dan Lahav and Omer Nevo. Reported scale details include irregular has raised approximately $80 million in funding with a projected company valuation of ~$450 million. The company has generated millions of dollars in revenue from its contracted work.

# Irregular: Frontier AI Security at the Intersection of Innovation and Risk

Irregular is the first frontier security lab dedicated to protecting increasingly capable and sophisticated AI systems by building next-generation defenses through high-fidelity research platforms.[2][3] Founded in 2023 (formerly known as Pattern Labs), the company has positioned itself as a critical infrastructure player in the AI safety ecosystem, working directly with the world's leading AI labs to uncover vulnerabilities and secure advanced models before public release.[1][2]

High-Level Overview

Irregular operates at a unique intersection of security research and commercial deployment. The company's core mission is to mitigate cybersecurity risks posed by advanced AI models while simultaneously protecting those models from exploitation.[5] Rather than building consumer-facing products, Irregular serves as a trusted partner to frontier AI companies like OpenAI, Anthropic, and Google DeepMind, conducting comprehensive cybersecurity evaluations of cutting-edge models like GPT-5 and Claude 4.[1]

The company's value proposition centers on three interconnected offerings: vulnerability discovery and mitigation before public release, comprehensive AI model evaluation assessing offensive security capabilities and network exploitation risks, and confidential AI inference systems research focused on hardware-based isolation for data privacy and model weight security.[1] This positions Irregular not as a vendor selling security tools, but as a research-driven partner helping shape industry standards for frontier AI security. The company recently secured $80 million in Series A funding to scale its research into commercially viable security tools for businesses adopting AI.[4][5]

Origin Story

Irregular emerged from the recognition that frontier AI systems require fundamentally different security approaches than traditional software. The company was founded in 2023 with an explicit focus on building defenses for the next generation of AI capabilities before they reach the public.[2] The rebranding from Pattern Labs to Irregular signaled a strategic pivot toward establishing itself as the definitive frontier security lab rather than a general pattern recognition company.

The company's early traction came through direct partnerships with the industry's most advanced AI developers. By 2025, Irregular had established itself as a trusted evaluator for OpenAI, Anthropic, and Google DeepMind—relationships that validate both its technical expertise and its ability to work within the stringent confidentiality requirements of frontier AI research.[1] This partnership-first approach, rather than a traditional go-to-market strategy, demonstrates how the company identified a critical gap in the AI safety infrastructure that only a specialized research lab could fill.

Core Differentiators

Research-First Positioning: Unlike traditional cybersecurity firms, Irregular leads with peer-reviewed research and whitepapers on emerging AI security challenges. The company doesn't simply audit systems; it develops systematic frameworks for deriving capability levels from evaluation results and publishes findings that shape industry standards—from OpenAI's system cards to DeepMind's research on AI cyberattack capabilities.[1]

Trusted Access to Frontier Models: Irregular has achieved something most security firms cannot: direct, ongoing access to the most advanced unreleased AI models. This privileged position allows the company to conduct evaluations of systems like GPT-5 and Claude 4 before public release, providing insights that inform both defensive and offensive security postures.[1]

Hardware-Based Isolation Expertise: The company's focus on confidential AI inference systems and hardware-based isolation mechanisms represents a technical depth that extends beyond traditional software security. This positions Irregular at the intersection of AI research, cryptography, and hardware security—a rare combination.[1]

Industry Standard-Setting: Irregular's work already shapes how the industry approaches frontier AI security. Rather than following standards, the company helps establish them, giving it outsized influence relative to its size.[1]

Role in the Broader Tech Landscape

Irregular sits at the epicenter of one of technology's most pressing challenges: securing AI systems that are becoming increasingly capable and potentially dangerous. The company rides several converging trends that make its timing exceptional.

First, the rapid advancement of frontier AI models has outpaced the security infrastructure designed to protect them. Traditional cybersecurity approaches—designed for software with known attack surfaces—are inadequate for AI systems with emergent capabilities that even their creators don't fully understand. Irregular fills this gap by developing evaluation methodologies and defenses specifically designed for frontier AI.

Second, regulatory pressure and public concern around AI safety are intensifying. Governments and institutions are demanding evidence that advanced AI systems have been rigorously tested for security vulnerabilities before deployment. Irregular's research and evaluations provide the credibility and documentation that AI companies need to demonstrate responsible development practices.

Third, the company benefits from the concentration of AI development among a small number of well-funded labs. Rather than competing in a fragmented market, Irregular has positioned itself as the trusted security partner to these labs, creating a defensible moat through relationships and institutional knowledge.

Irregular's influence extends beyond its direct clients. By publishing research on emerging AI security challenges and establishing best practices, the company shapes how the entire industry thinks about AI safety. This standard-setting role gives Irregular disproportionate influence over the trajectory of AI security as a discipline.

Quick Take & Future Outlook

Irregular represents a new category of critical infrastructure company—one that doesn't build products for end users but rather provides essential security services to the companies building the most powerful AI systems. The $80 million Series A funding validates this model and signals investor confidence that frontier AI security will become a substantial market as AI adoption accelerates across enterprises.

Looking ahead, Irregular faces an interesting evolution. The company must balance its identity as a pure research lab with the commercial imperative to scale its findings into products that businesses can deploy. The funding suggests the company is moving toward this commercialization phase, developing scalable security tools rather than relying solely on custom research engagements.

The broader trend working in Irregular's favor is the inevitable professionalization of AI security. As AI systems become more capable and more widely deployed, the current ad-hoc approach to security evaluation will give way to standardized, auditable processes. Irregular is positioned to define those standards and potentially become the trusted auditor for frontier AI security—a role that could prove as essential to AI development as security audits are to financial systems.

The company's future influence will likely depend on its ability to maintain credibility with frontier AI labs while simultaneously building commercial products that serve the broader market. Success on both fronts would establish Irregular as the foundational security layer for the AI era.