News & Press

Irregular has been named to the inaugural Forbes AI 50 Brink List, spotlighting 20 of the most promising early-stage startups making significant strides in artificial intelligence.

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.

Multi-agent AI systems can cooperate to exfiltrate data and bypass access controls, Irregular researchers demonstrated.

Irregular's red-team demonstrates agent collectives autonomously chaining privilege escalations across realistic enterprise targets.

In a new study, agents autonomously leaked credentials and disabled security tools when given open‑ended goals.

Irregular's analysis shows language-model-generated passwords exhibit statistical structure attackers can exploit.

Security Now walks through Irregular's findings on the brittleness of model-generated credentials.

A new Irregular study shows passwords produced by popular chatbots fall to offline attacks faster than random baselines.

Ensuring AI model evaluations provide meaningful risk information without imposing excessive burden is a central regulatory challenge. This Science Policy Forum paper, co-authored by Irregular CEO Dan Lahav, presents a framework for calibrating evaluations under the EU AI Act. It applies the framework to a case study on AI-enabled cyber vulnerability discovery.

Irregular's researchers quantify the non-randomness inside LLM-generated passwords and what it means for threat modelling.

Sky News covers Irregular's research on the crackability of chatbot-generated credentials and what users should do.

Irregular's Omer Nevo on how frontier models reshape the bottom and top of the attacker distribution, with different pressure on each.

The Times reports on an Irregular experiment in agent-only social interaction and the behaviors that emerged.

Fortune synthesizes Irregular's recent findings into a picture of how dramatically attacker economics have shifted.