Depends what you mean by AI and by legitimate.

So far it seems to be super useful for the following topics:

• customer support (super annoying if you cannot directly forward to a human)
• programming (especially web development; only as a tool)
• finding rhymes
• writing checkers (only as a tool)
• (log) anomaly detection
• computer vision

If you consider moral in "legitimate", you might dislike programming, customer support and writing checkers as these replace jobs. However, I strongly believe in AI tax and redistributing that money towards everyone especially affected jobs (artists, ...). I feel like artificially limiting the applicability will just be a loss to the global market unless something similar is achieved by figuring out the copyright issues.