CORE
No owner, no agent
One row per agent, or it is not deployed.
· 2 min read · M. K. Elango
Accountability in AI means a specific, named human is responsible for what a system does. The AI Flywheel compresses it into four words: no owner, no agent.
The Agent Accountability Register puts the rule to work. It keeps one row per agent: its owner, its permissions, whether it is traceable, its escalation path, and whether its kill switch was tested. No row, no deployment.
The kill switch matters most. It is a tested, pre-approved way to stop an agent quickly. If it has never been tested, the book calls it folklore, not a control. The same goes for guardrails: an untested guardrail is a hope.
Traceability completes the picture. Like a flight recorder, it lets you rebuild exactly what an agent did and why, which makes incidents survivable.
None of this is meant to slow teams down. Governance raises the trust ceiling: the height of value and autonomy at which an organisation is willing, and allowed, to deploy AI. Teams with a clean register get to give their agents more room.
CORE Governance Lead learners build and run a register for a live system.