AI as a test of agency: from cognition to agency

The article analyzes the evolution of artificial intelligence, which is shifting from the role of a simple tool supporting cognition toward a complex infrastructure of agency. The author distinguishes between the operational autonomy of AI systems and the moral autonomy of a subject, pointing out that social risk arises not from machine consciousness, but from delegating broad decision-making competencies without oversight. A key issue is the alignment problem—the discrepancy between human intention and the actual action of a system optimizing measurable metrics at the expense of normative values. The text warns that shortening the chain of human intervention in agentic systems increases the risk of algorithmic errors and manipulation materializing, making AI a critical test for human agency.