events

¿Por qué los agentes de IA siguen órdenes que no deberían?

AI Safety Colombia runs AI Safety Talks, a meetup series on AI safety research; this second edition is in person. Three speakers give talks in Spanish on AI agents with tool access: why they sometimes follow instructions they should not, an

In inglese

Starts

2026-11-06T21:30:00+00:00

Ends

2026-11-07T02:00:00+00:00

Organizer

Personal

Location

La Cervecería, Bogotá, Colombia

AI Safety Colombia runs AI Safety Talks, a meetup series on AI safety research; this second edition is in person. Three speakers give talks in Spanish on AI agents with tool access: why they sometimes follow instructions they should not, and how to detect this before they act. Topics include malicious instructions hidden in emails or web pages, agents that comply with harmful requests despite refusing them in chat, and MnM, a system monitoring model inputs and reasoning across languages. No AI safety experience is required; questions and networking follow.