# Luis Cosio

> Hardening frontier AI systems against nation-state adversaries and loss-of-control failure modes.

I work at the intersection of frontier AI and high-security systems. My focus is turning abstract safety and security requirements into deployable technical artifacts that can withstand real adversaries, from nation-state attacks to loss-of-control scenarios.

## Focus Areas

- Frontier AI Security (SL5): Securing model weights, inference systems, and lab infrastructure against nation-state threats and loss-of-control failure modes.
- AI Alignment and Interpretability: Mechanistic interpretability, scalable oversight, and monitoring-based interventions.
- Security Engineering: Air-gapped deployments, attack surface minimization, and high-assurance runtime design.

## Blog

- [Distillation Caltrops](https://www.luiscos.io/blog/caltrops/): A watermark for unauthorized distillation that lives in a learned reasoning habit. The teacher uses the habit well; a distilled student inherits the form without the substance, fails confidently, and leaves a detectable fingerprint that survives paraphrase.
- [μInference](https://www.luiscos.io/blog/uinference/): From minimal stack to SL5 Weight Enclave. seL4-backed prototype for frontier model weight protection.
- [OpenWild](https://www.luiscos.io/blog/openwild/): A new conversational dataset for modern language models, inspired by WildChat.
- [Mechanistic Watchdog](https://www.luiscos.io/blog/mechanistic-watchdog/): Real-time cognitive interdiction for LLMs.

## Selected Work

- [Security Level 5 Task Force](https://sl5.org/projects): SL5 security artifacts for frontier AI systems.
- [MechWatch](https://mechwatch.luiscos.io/): 1st place, Defensive Acceleration Hackathon.
- [MATS Mentor](https://www.matsprogram.org/): Mentoring scholars in the ML Alignment & Theory Scholars Program.
- [SPAR Mentor](https://sparai.org/): Mentoring researchers starting careers in AI safety and alignment.
- [Algoverse Mentor](https://algoverseairesearch.org/): Guiding student research teams in AI safety.

## Background

Mexico City based technologist with 15 years across cloud, cybersecurity, and AI. Currently working full-time on AI safety and security.

## Connect

- Website: https://www.luiscos.io
- Twitter: https://twitter.com/luiscosio
- LinkedIn: https://www.linkedin.com/in/luiscosio/
- GitHub: https://github.com/luiscosio
- Google Scholar: https://scholar.google.com/citations?user=BbcHU5YAAAAJ

Last updated: 2026-06-09