liveBook · Manning

Cascading failure

As explained in AI Governance

A cascading failure is a sequence in which an error or compromised output propagates through connected agents and contributes to further failures.

A cascading failure occurs when an error or compromised output propagates through connected agents and contributes to further failures. section 4.8.1

Treat it as a distinct security risk of autonomous AI systems, alongside tool misuse, memory poisoning, and rogue agents. section 4.8.1

How AI Governance covers it

How deep it goes
Defines propagation through connected agents and names the risk, but does not explain its mechanisms or controls.
What you can do after
Understand cascading failure as a distinct security risk in autonomous AI systems.
How it connects
Connects cascading failures with tool misuse, memory poisoning, and rogue agents.

Also explained in

Compare how 2 books explain cascading failure →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of AI Governance. The section links point to the parts of the book it draws on.