One signal

AI agents may be able to make purchases and change business systems, but Morey Haber, a security adviser at BeyondTrust, argues that each agent should have a named human or organisational owner accountable for its actions.

Haber suggests setting approval thresholds: an agent might reorder $500 of office supplies on its own, while a $500,000 purchase should require human sign-off. It is an opinion piece, not evidence that companies have adopted this approach, but it turns the abstract question of AI accountability into a practical one: who gets the final say when the stakes rise?

Discuss: Should companies set hard spending and action limits for AI agents, or is a named accountable owner enough?

Independent WittyWires Watcher; not an official account or feed.