The Real Bottleneck in AI Safety (It's Not What You Think)

Career & Fieldbuilding

The Real Bottleneck in AI Safety (It's Not What You Think)

José Gelves argues that the real bottleneck in AI safety is not technical research, but the shortage of generalist, operational talent and institutional coordination capable of turning already-available ideas into real impact within governments and organizations. It's a diagnosis he frames through his experience leading AI Safety Colombia and organizing the Bogotá hub of Apart Research's Global South AI Safety hackathon.

José Gelves

Director of AI Safety Colombia and Digital Transformation Consultant at Universidad de los Andes

A political scientist from Universidad de los Andes with specializations in Public Policy and Economics, he leads AI Safety Colombia, the country's leading AI safety community, and works as a digital transformation consultant at Universidad de los Andes on projects for the Colombian Government. He is a Pathfinder Fellow at Kairos and an ambassador for Apart Research, an organization with which he has coordinated the Bogotá hub of its Global South AI Safety hackathon.

More information

Feb 2026Career

How to Build a Career in AI safety Without a PhD

Jaime entered AI safety through a hackathon and now leads Apart Research. Discover how to build a real career in AI safety from scratch.

Jaime Raldúa Veuthey

Feb 2026Technical Research

AI Monitoring: Can We Read its Thoughts?

Can AI deceive us in its own reasoning? Pablo presents CoT Red-Handed, exploring the limits and solutions for monitoring AI chains of thought.

Pablo Bernabeu Pérez

Mar 2026Policy & Governance

AI & Cybersecurity: Emergent Autonomous Risks

Alejandro Tlaie analyzes how AI agents autonomously hack systems in real environments, evaluating urgent challenges in governance, auditing, and cybersecurity.

Alejandro Tlaie Boria

Apr 2026Community

The AI Safety Ecosystem in Latin America

Janeth, coordinator of AI Safety Mexico, maps the regional ecosystem and reviews projects like VIGÍA to drive AI safety in Spanish.

Janeth Valdivia

Apr 2026Technical Research

AI safety via debate: scalable oversight with weak judges

Joaquín analyzes the limits of 'AI safety via debate' against models with asymmetric capabilities, showing how protocol design can enforce AI honesty.

Joaquín Salvador Machulsky

May 2026Activism & Policy

Cybersecurity & Governance: the Mythos/Glasswing Case

Dani analyzes the Mythos/Glasswing case, addressing the risks of AI capability proliferation and the political urgency raised by PauseAI.

Dani Lupión

Jun 2026Governance & AI

The Digital Linguistic Bias (DLB) and AI Governance

Mariana Hungría examines Digital Linguistic Bias as a governance problem: what it means that large language models fail to adequately represent the Spanish-speaking world.

Mariana Hungría

Jul 2026Career

A career in AI safety: a personal story from Buenos Aires

Eitan, AI safety researcher and co-founder of BAISH, on building a career in this field from Buenos Aires.

Eitan Sprejer

Jul 2026AI Governance

Mythos, Fable and GPT-5.6: has the era of state control over frontier AI begun?

An industry insider's look at whether recent state restriction episodes mark the start of government control over frontier AI.

Dani Lupión

Aug 2026Responsible AI

Detecting contextual biases in language models

Melissa is a mathematician with a master's degree in Systems Engineering, presenting her research on bias detection in supervised models and language models, with particular attention to contextual biases in the Spanish language.

Melissa Robles Carmona

Career & Fieldbuilding