Copyright in AI Outputs: Who Owns AI-Created Works?

Why It Matters

Businesses must navigate new AI disclosure rules and limited copyright scopes, or risk losing protection for AI‑assisted works while ensuring innovation remains unhindered.

Key Takeaways

• Courts reject AI‑generated works as copyrightable without human authorship.
• Copyright Office now requires AI disclosures and limits protection to human contributions.
• Registrations can cover only selection, arrangement, or translation of AI outputs.
• Over‑protecting AI content could lock down public domain and hinder innovation.
• Legal consensus treats AI as tool, not author, preserving human‑centric copyright.

Summary

The Harvard Law panel examined whether AI‑generated creations can qualify for copyright protection, highlighting recent litigation and evolving Office guidance.

Judges consistently ruled that works produced autonomously by machines lack the requisite human authorship, as illustrated by Stephen Taylor’s rejected registration and the Supreme Court’s denial of cert. The Copyright Office now mandates explicit AI disclosures and limits registration to the human‑originated elements such as selection, arrangement, or translation.

Panelists cited the “Azari the Don” graphic novel case, where only the curated compilation earned protection, and drew parallels to historic photography standards. Rebecca Tushnet emphasized the human‑threshold rule, while Jessica Silbey described the Office’s emerging “anti‑prompt” policy that bars pure AI outputs from copyright.

The consensus warns that extending full copyright to AI‑generated material could flood the public domain with locked‑down content, stifling downstream innovation. Balancing modest incentives for human creativity with open‑access principles will shape future IP strategy for creators, studios, and tech firms.

Original Description

Generative AI can produce striking images and polished text in seconds, but copyright still turns on a basic question: Did a human author do enough? At a panel discussion on April 22, Harvard Law Professors Christopher Bavitz and Rebecca Tushnet along with BU Law School Professor Jessica Silbey discussed the copyright issues implicated by Generative Artificial Intelligence tools and training of large language models, focusing on the copyright status of AI outputs.