Funding to harden model security for high-capacity AI
AI security startup Irregular announced an $80 million funding round led by Sequoia and Redpoint that the company says will expand its work protecting “frontier” models from misuse. The startup plans to scale detection, red-teaming and runtime controls that aim to prevent malicious exploitation of large models as capabilities grow rapidly.
Company executives said investors are backing a market they view as essential to safe deployment of powerful AI systems. Observers note the deal reflects rising investor interest in tools that monitor and govern advanced models as regulators and firms push for stronger safety measures.

Productising model safety at scale
Irregular is positioning its stack for enterprise customers and cloud providers, offering APIs and integration that monitor model inputs and outputs for anomalous or risky behavior. Analysts said demand for such services is increasing as more organizations deploy custom, high-capacity models with broader attack surfaces.