Agents and agency

When an AI agent performs a task, who is responsible? This is a question we have to grapple with now for the first time.

The answer should be clear, though already the waters are getting muddied: a human always holds responsibility. One human. Every time.

AI is an extraordinarily powerful tool. It is gasoline on the fire of human intelligence and agency. Absent those ingredients, it is inert. We should not pretent otherwise.

An AI agent is not an agentic entity unto itself. It is computer software. It is made of math. Data in, data out.

A coding agent is no more responsible for a line of code than a word processor (or a pen) is for a line of prose. The coding agent, word processor, and pen are tools. A human is the creator. That human is responsible.

As we hurtle forward in this time of great change, this is a principle we must keep clear and top of mind.