Security
Metorial blog posts about security.
Security
The first hour after an AI agent makes a damaging change
A practical response sequence for containing an agent, preserving evidence, finding the full scope, and recovering without losing the trail.
Ivan SiianytsiaSecurity
When should an AI agent ask a person before it acts?
A useful approval rule starts with the action’s impact and reversibility, not a blanket “human in the loop” requirement.
Tobias HerberSecurity
Security has to come first
AI agents are starting to touch real company systems, which means security cannot be something added later. This is how we think about isolation, key control, and the protocol layer at Metorial.
Tobias Herber