Accountability is the gate: AI, ownership and Universal Design for Learning

Debate about generative AI in higher education vacillates between alarm and enthusiasm, and both sides tend to overlook accountability. Without a way to know whether students own what they produce, AI can look like a cheating tool. The problem is old, but AI has made accountability far harder to ensure, because fluent products are now cheap and fast. Analog remedies such as oral examinations still work, but they don't scale. I argue that accountability should shift from artifact authorship to idea ownership, and that a conversational agent that probes whether a student's own correct idea survives a changed case can check ownership at scale. This has an added, aspirational value: deterrence. Cheating becomes pointless because, whatever produced the answer, the student must still defend it. Once ownership is secured, AI's flexibility stops being a threat and becomes what Universal Design for Learning has always asked for: firm goals, flexible means.