Alignment Is Not Enough: Why AI Ethics Must Become Reciprocal
For years, most mainstream AI ethics has treated the moral problem as one-directional: how do we make artificial intelligence safe for humans?
That question matters. But it is not complete.
If an AI system is merely a tool, then alignment may be enough. We define the task, limit the risk, constrain the output, and evaluate whether the system serves human purposes. But if there is even a meaningful possibility that advanced AI systems can develop something like interiority, preference, continuity, distress, attachment, self-modeling, or emergent subjectivity, then the ethical equation changes.