training data sourced without consent vs the artistic legitimacy of work no human authored

AI Art Ethics — Who Owns Work the Machine Made?

Generative models can produce paintings, music and prose at near-zero cost. The training data, in many cases, was scraped from artists who never consented and were never paid. The legal answer is unsettled across jurisdictions, but the ethical question is sharper: if the model learned a living artist's style and a stranger now sells outputs in that style, what does authorship even mean? And when the output is technically new but recognisably derivative, who bears the moral cost — the user, the model maker, or the platform that hosts it?

Vote on this dilemma

An AI generates a masterpiece painting with no human creative input. Who owns the copyright?

The AI's creator companyNo one — it should be public domain

Vote now →See full results →

Research background

Ongoing court cases in the US and EU are testing whether training a model on copyrighted work counts as fair use or as infringement. The Stanford CRFM and the EU AI Act both treat training-data provenance as a first-class concern; the moral question of stylistic imitation runs ahead of the legal one.

SplitVote is for entertainment and aggregate insight, not a scientific test.

You have a deepfake video of a public figure committing a crime. You don't know whether the real event happened — only that the video itself is fake.

Release itDestroy it

AI will eliminate 30% of all jobs in 10 years. Governments can slow it down at massive economic cost, or let it happen and retrain workers.

Slow it downLet progress happen

A self-driving car's brakes fail. It must choose: swerve into a barrier (killing the passenger) or hit a pedestrian who jaywalked.

Hit the barrierProtect the passenger