Yuval Noah Harari made a useful distinction at Davos: AI is not a tool in the way we have traditionally understood tools. A knife does not decide what to cut. An AI system can interpret an objective, select an action, and generate language intended to influence a person.
Agency changes the responsibility
The more freedom a system has to choose the steps between an instruction and an outcome, the more carefully its context and boundaries must be designed. Accuracy testing alone is not enough. Teams also need to examine manipulation, hidden assumptions, escalation paths, and who remains accountable.
Ethics belongs inside the work
For teams building or deploying AI, ethical questions are not a separate philosophical exercise. They affect architecture, data access, review, user experience, and release decisions. Staying current on those questions is part of the job.