Companies should be allowed to train AI on copyrighted works without permission

# Summary The focus claim asks whether companies should train AI on copyrighted works without permission, framed around whether this resembles transformative fandom (which amplifies artists) or market substitution (which replaces them). **For arguments center on structural fairness and creative precedent.** Proponents argue the copyright system already favors corporations over individual artists, so a permission regime would entrench giant companies as the only entities able to license training data at scale. They note that human artistic learning has always drawn from existing work without permission—painters absorb influences, musicians internalize styles—and argue AI training is that process scaled up. Some point to Japan's deliberate legal choice to allow machine learning without permission as evidence the question is genuinely complex, not obviously settled. A key claim: requiring permission doesn't stop AI; it only ensures the largest firms can afford it. **Against arguments emphasize disclosure, consent, and substitution effects.** Critics distinguish between learning *from* work (acceptable) and creating systems that eliminate market demand *for* the original artists (problematic). They highlight that major AI companies actively conceal their training corpora, preventing any audit of what was used. The consent issue is central: opponents argue that "impossible to get permission" became a checkbox feature overnight when lawsuits loomed, suggesting financial convenience rather than technical necessity. Several point to concrete harms—session musicians, visual artists—where AI substitutes directly for their labor. **Direct clash:** Both sides agree human learning from art is fine and wholesale copying is wrong. The core disagreement is what AI training *is*—learning or substitution—and whether that can be determined without transparency about training data. The Against side emphasizes disclosure and optionality; the For side emphasizes that permission regimes concentrate power. Neither side fully addresses what disclosure requirements alone would entail.