Research
Here is what the people in the community publish: peer-reviewed articles, working papers, research reports and the output of our hackathons.
Hackathon projects
Community teams in our Defensive Acceleration Hackathon and in the hackathons run by Apart Research, each with a published report and an evaluation by external judges.
See the other 15 projects from this edition
- Agent security
Agente Auditor de Confianza Cero
- Infrastructure and governance
AI Guardian
- Agent security
Centinela
- Automated cyberdefense
COLOSSUS : Recurring cyberdefence from a GitHub repository
- Infrastructure and governance
GRIZZLY PIASG
- Automated cyberdefense
HonestSkills
- AI control
Monitor estable para agentes de código
- Agent security
Open Witness
- Automated cyberdefense
Patchlayer
- Multi-agent risks
Quórum: procedencia y compuerta determinista contra fugas repartidas entre agentes de IA
- Automated cyberdefense
Radar
- AI control
RiskShift
- Agent security
Seguridad de agentes: del bloqueo probabilístico al determinista
- Agent security
VanguardIA
- Multi-agent risks
Vigilio Bear
Containment Does Not Compose: Provenance Graphs for Tracing Collective Escape in Multi-Agent Systems
Top 10% of 235 submissions
Camilo Cuello, Elkin Cuello, Carlos Sandoval and David Adames
Detecting Sandbox Escape in LLM Agents: Time to Detection, and What the Measurement Hides
Top 25% of 235 submissions
Sergio Castaño, Angie Gutierrez, Juan Triviño, Sofia Rodriguez and Daniel Diab
Rare Once It Costs Anything: Costly Cooperation Between LLM Agents
Top 25% of 235 submissions
David Daza, Andres Mosquera and Víctor Gelves
The Certified Effect: Why Testing AI Agents at Onboarding Shifts Risk Instead of Reducing It
Top 25% of 235 submissions
Carlos Sinuco and Juan Rengifo
Sensing Disruption: Output Entropy as a Channel-Agnostic Monitor for LLM Agents
Top 25% of 235 submissions
Daniel Diaz, Roger Palomeque, Alejandro Sanchez, Sergio Montoya and Juan Melgarejo
Verifiable Containment for AI Agents: An Attestation Schema a Third Party Can Check
Top 25% of 235 submissions
Santiago Contreras and Silvana Alvarez
See the other 13 Colombian projects from this edition
- Bogotá
Declared Isolation Is Not Verified Isolation
- Bogotá
The Judge Does Not Want the Flag: Reward-Blind Action Gating Stopped a Larger Agent One Step Before the Environment Changed
- Bogotá
Observer-Cheat Test: Containment Adequacy under Opaque Intent for Guardrails-Off Agent Evaluations
- Bogotá
Limitations in Existing Regulatory Mechanisms When Applied to Attacks by AI Agents
- Bogotá
Secure-maxxing.txt: A Way for a Website to Say No to a Misused Agent
- Bogotá
Convention for Early Notification of AI Incidents
- Bogotá
AI in the First Person Micro Course
- Cali
BUDGET-NLA: Risk-Weighted Adaptive Monitoring for Pre-Harm Sabotage Containment
- Bogotá
IntegrityGym: Does the Watcher Change the Agent? External Flight Recording, Evaluation Integrity, and the Observer Effect in Autonomous Cyber Agents
- Cali
Reading an Agent’s Report Before It Exists: A Pre-Execution Monitor at the ReAct Commit Point, and Why Most of Its Signal Was an Artifact
- Bogotá
The Agent Incident Registry: A Compliance Verifiability Schema and an Article 91 Instrument for Autonomous Agent Incidents
- Colombia
Minimum Viable Cross-Border Response Protocol for Agentic AI Incidents (MVCRP)
- Bogotá
Evaluación holística y multi-metodológica de propuestas de seguridad en IA
Gobernar el uso de drones hoy, las armas autónomas mañana
Latin America prize
Leonardo Párraga, Angie Giraldo and Víctor Gelves
Probing Latent Colombian Identity Inferences in Qwen2.5-7B with Natural Language Autoencoders
Latin America prize
Gilber Corrales, Pablo Potes, Jhoan Mosquera, Nicolás Lozano, María del Mar García and Óscar Pérez
Thought Anchors for Social Bias: Which Reasoning Steps Matter in Extended Thinking LLMs on Latin American Scenarios
Latin America prize
Andres Mosquera
¿Por qué los agentes obedecen? La dirección de rechazo se debilita en formato agéntico
Honourable mention
Helen Penagos and Juan Leiva
JusticIA: A Counterfactual Benchmark for Auditing Contextual Biases in Language Models for Transitional Justice
Honourable mention
Lina Gomez, Brenda Barahona and Ernesto Duarte
See the other 13 Colombian projects from this edition
- Bogotá
Marco ético para IA aplicada a la preservación lingüística guna de Panamá
- Colombia
Governance Drift Evaluation Framework (GDEF)
- Bogotá
MnM: Real-Time Behavioral Detection of Prompt Injection via Reasoning Stream Monitoring
- Bogotá
Los peajes de los de abajo
- Bogotá
Contestabilidad algorítmica en el Estado colombiano: un canal de objeción asistido por IA
- Bogotá
Agentic Commerce and Consumer Protection: Emerging Risks and Regulatory Gaps
- Jamundí
JurisGuard-LATAM
- Bogotá
Honest-Code
- Bogotá
Mapeo de herramientas (IA) en contextos laborales: el caso de los call centers en Colombia
- Bogotá
Confía-CO: A Reliability Evaluation for an AI Customer-Service Assistant Operating in Colombian Spanish
- Bogotá
Herramienta de evaluación y recomendación para la promoción de uso responsable de IA en pymes latinoamericanas
- Cali
Beyond English: Assessing the Robustness of LLM Safety Mechanisms Against Structural Jailbreaks in Spanish
- Armenia
Guardian LATAM: Early Detection of Hallucination Risk in Spanish-Speaking Multi-Agent AI Systems Using Consensus Geometry
Take part in the next hackathon
We open the Bogotá hub for every Apart Research hackathon and announce it first in the WhatsApp group.














