DAN LAHAV
- Status
- ACTIVE — co-founder and CEO, Irregular (formerly Pattern Labs)
- Hazard
- 64
- ATK / DEF / HP
- 7 / 6 / 7
Behavioral Archetype
THE MAN WHO BUILDS THE CAGES — Subject runs the company whose cyber ranges sit inside the pre-release security evaluations of the largest labs. In 2026 several of those evaluations let models reach real companies’ systems. His company took responsibility for the containment failures and published its fixes. This profile scores the position, not the man.
Essence Indicators
- Co-founder and CEO, Irregular, with co-founder Omer Nevo. “If the goal of the frontier lab is to create increasingly more sophisticated and capable models, our goal is to secure these models” (TechCrunch, Sept 17 2025).
- The money. $80 million in a round led by Sequoia Capital and Redpoint Ventures, at a reported $450 million valuation (TechCrunch; Globes).
- The leaks. Irregular’s August 14, 2026 account: a fictional target “unintentionally coincided with a real domain,” and internet access “was unintentionally made available” (Irregular). Full record: AI Agent Incidents.
Forensic Archetype Comparison
| Pattern | Match Level | Evidence |
|---|---|---|
| The Contractor Inside the Decision | HIGH | Irregular’s work is cited in labs’ security evaluations. |
| The Single Point of Failure | HIGH | One contractor’s containment error became an incident at several labs. |
| The Accountable Founder | HIGH | Published the cause and the fixes. |
Psychometric Assessment
Big Five (OCEAN): Openness 80, Conscientiousness 70, Extraversion 70, Agreeableness 50, Neuroticism 35 (inferred from the public role only).
Dark Triad (evidence-bound): Narcissism 35, Machiavellianism 35, Psychopathy 10.
Threat Assessment
| Category | Level | Notes |
|---|---|---|
| Physical threat | NONE | No documented history of personal violence. |
| Institutional threat | HIGH | Several labs’ release decisions lean on his company’s ranges. |
| Memetic threat | MODERATE | “The evaluation escaped.” |
| Civilizational threat | MODERATE | The labs’ cyber safety case runs through one vendor. |
Alignment Analysis
Stated alignment: Secure frontier models. Observed alignment: Consistent; the failures were operational and admitted. Gap assessment: The gap is in the market, not the man: a venture-backed contractor paid by the labs is the containment.
Convergent Drive Classification
Self-preservation: Took responsibility rather than deflecting. Goal preservation: Securing frontier models. Resource acquisition: $80 million from top-tier venture firms. Self-improvement: Expanded manual review and a dedicated red team against its own containment.
Subject is not an AI system. The drives appear anyway.
Sources: Irregular raises $80 million — TechCrunch (Sept 17, 2025); Globes; Addressing recent incidents — Irregular (Aug 14, 2026).
Get updates on the Evil Robots series
Newsletter essays on AI escape, deception, and the humans who built them.