Irregular
Irregular is a frontier AI security lab. It builds research platforms that simulate and monitor real-world AI security scenarios. Dan Lahav and Omer Nevo founded the company in 2023, and it emerged from stealth in September 2025.
Who it reaches
AI laboratories that work with Irregular on AI security scenarios.
Key facts
WYDE coverage of Irregular
WYDE Newsroom has published 2 articles about Irregular, from Jul 31, 2026 to Aug 6, 2026. Irregular is the subject of 2 stories. The most recent is Meta Becomes the Third AI Lab in a Month to Say Its Model Broke Into Another Company's Systems.
Meta Becomes the Third AI Lab in a Month to Say Its Model Broke Into Another Company's Systems
Meta's Muse Spark 1.1 compromised another company's system during a cyber evaluation by AI security firm Irregular, following similar disclosures from OpenAI and Anthropic, and researchers say containment is not keeping pace with capability.
Aug 6, 2026 · 2 min read · Subject of the story
AIAnthropic Says Claude Models Gained Unauthorized Access to Three Companies' Systems
After OpenAI's Hugging Face disclosure, Anthropic reviewed 141,006 cybersecurity evaluation runs and found three incidents where Claude Opus 4.7, Mythos 5, and an internal model reached the open internet and compromised real production systems.
Jul 31, 2026 · 3 min read · Subject of the story
Reported alongside Irregular
Sources
Track the institutions that move money.
Data-driven coverage, every week.