I don't think that is possible. Humans have always taken the path of least resistance, especially when it comes to work/school.

The idea that we just "trust everyone to carefully check and learn from AI output" as our barrier to human skillsets eroding is never going to work.

There is an Anthropic engineering post on HM front page that addresses this exact issue:

"... supervise the agent’s behavior via a human-in-the-loop. Claude Code previously protected against agents taking unintended actions by asking users for permission at each turn. Theoretically that works, but we’ve found the approach to be fallible. Our telemetry showed users approved roughly 93% of permission prompts. The more approvals a user sees, the less attention they pay to each, becoming over time much less diligent in their supervision. "