Series A: USD 80 million for AI cyber security specialist Irregular

In a market where the intro­duc­tion of AI is progres­sing rapidly, the Israeli startup Irre­gu­lar has now closed a finan­cing round of 80 million US dollars. The inves­tors include Sequoia Capi­tal, Redpoint Ventures, Swish Ventures, Assaf Rappa­port and Ofir Ehrlich. The company, foun­ded by Dan Lahav and Omer Nevo, specia­li­zes in stress test­ing advan­ced AI models under real-world condi­ti­ons, working with leading AI labs such as OpenAI and Anthropic. 

Irre­gu­lar was foun­ded in 2023 by Dan Lahav and Omer Nevo. Lahav, who previously worked at LabPi­xies, a startup acqui­red by Google, and later worked as an AI rese­ar­cher at IBM, brings exten­sive expe­ri­ence in AI rese­arch. Nevo is a serial entre­pre­neur with a back­ground in cyber­se­cu­rity. Toge­ther, they have built a company that focu­ses on evalua­ting and secu­ring advan­ced AI models. The new funding will be used to expand rese­arch and deve­lo­p­ment, expand the team, inten­sify colla­bo­ra­tion with leading AI labs 

Busi­ness model

Irre­gu­lar’s busi­ness model is based on conduc­ting stress tests for AI systems under real threat scena­rios. This includes tests that focus on how AI models react to attempts to circum­vent anti­vi­rus soft­ware or carry out auto­no­mous offen­sive operations. 

The aim is to iden­tify poten­tial vulnerabi­li­ties and deve­lop solu­ti­ons to ensure the secure imple­men­ta­tion of these systems on a large scale.

The foun­ders explain: “It’s about prac­ti­cal tools that stop thre­ats where­ver AI is deve­lo­ped or used.”

The start-up’s tech­no­logy has alre­ady attrac­ted the atten­tion of leading AI labs: OpenAI uses Irre­gu­lar’s ratings in system maps for GPT‑3, GPT‑4 mini and GPT‑5. Irre­gu­lar has also co-autho­­red a white paper with Anthro­pic on confi­den­tial infe­rence systems, while rese­ar­chers at Google Deep­Mind have cited the company’s plat­form in studies on AI cyber­at­tack capabilities.

The funding reflects a broa­der realiza­tion across the tech­no­logy land­scape that the success of AI depends not only on buil­ding ever more powerful models, but also on protec­ting them from misuse, acci­dents and new threat vectors. As compa­nies and govern­ments acce­le­rate adop­tion, the demand for proac­tive AI secu­rity has increased. With its strong inves­tor support and early trac­tion, Irre­gu­lar is posi­tio­ning itself as a funda­men­tal player in this emer­ging layer of AI infrastructure.