Vai al contenuto
AI.info

events

When Agents Go Rogue: What We Can Learn from recent AI Incidents

Un incontro del Toronto Java Users Group sui recenti incidenti che coinvolgono agenti di IA, sui controlli relativi agli agenti LLM e sulle domande da porsi prima della messa in produzione.

When Agents Go Rogue: What We Can Learn from recent AI Incidents | Eventi sull’IA

In inglese

Starts
2026-11-26 23:30 UTC
Organizer
Toronto Java Users Group
Location
Free Times Cafe, 320 College St, Toronto, Canada

Toronto Java Users Group hosts a talk by Mohamed Mansour, a principal engineer and architect, on recent AI agent incidents and what they reveal about agent design. The session starts with how an LLM becomes an agent, then examines the surrounding harness: tools, instructions, skills, permissions, hooks, human approval and monitoring. It covers MCP, Agent Skills and A2A, maps reported incidents to missing controls, and discusses questions to ask before deploying agents in production. No prior experience with AI agents is required.

Event website