Cristiano Luchini

I evaluate frontier LLMs for Scale AI, Mercor, Invisible, and Labelbox, conducting safety-critical annotation, adversarial red teaming, and complex RLHF evaluations in English and Italian.

My Data Trauma Theory examines how pathological pre-training data can produce incoherent internal world-models and unsafe behavior under stress. I executed 6,000+ high-intensity experiments across 17 frontier LLMs and designed a mass-context-injection pipeline using 99 structural code pathogens.

I also developed Preventive Chaotic Intelligence, OSR-Spanda, and Custos AI to map pre-deployment risk, stabilize agent behavior, and forecast systemic governance failures. This work was selected for the IRCAI Global Top 100 under the auspices of UNESCO.

Alongside AI safety research, I've spent more than 12 years as an Editor-in-Chief and investigative columnist, publishing 900+ analytical pieces on AI, digital ethics, philosophy, and socio-technical change. My background in artistic direction, filmmaking, and systems design informs how I investigate the collision between human consciousness and machine logic.