How to File Taxes in Canada (2025): Step-by-Step CRA Guide for Beginners
In recent years, AI image generation tools like DALL·E and Midjourney have enabled anyone to create sophisticated visuals from textual prompts. But alongside their growth comes a thorny legal question: who owns such images, and can they be protected under copyright law? This article surveys the current landscape of legal doctrine, platform terms, and ongoing litigation related to AI-generated art in 2025.
Traditional copyright systems (e.g. U.S. Copyright Act) generally require a “human author” to qualify for protection. The U.S. Copyright Office has reaffirmed that works produced solely by AI, without significant human creative input, do not meet the authorship requirement. (copyright.gov)
In March 2023, the U.S. Copyright Office published guidance titled “Works Containing Material Generated by Artificial Intelligence”, stating that AI-generated elements must be distinguished from human creative contribution, and pure AI output is not registrable. (copyright.gov)
Courts have reinforced this boundary: in March 2025, a U.S. appeals court declined to grant copyright to an AI-generated artwork for lack of human authorship. (reuters.com) Moreover, a Czech municipal court held that a DALL·E-generated image cannot be a copyrightable work under Czech law, given its non-human “authorship.” (ipkitten.blogspot.com)
OpenAI’s terms historically granted users rights to use, reproduce, and sell images generated via DALL·E, subject to content policy. (medium.com) However, those terms also contain clauses assigning any rights the user might have to OpenAI. (findlaw.com) In practice, the “ownership” is limited: even if you can sell or use the image, OpenAI retains broad rights to reuse or license it.
For DALL·E 3, OpenAI has implemented constraints: it declines requests to explicitly replicate the style of living artists, to reduce infringing outputs.
Midjourney’s Terms of Service require users not to infringe third-party IP and disclaim liability for misuse. (docs.midjourney.com) Its licensing model grants users varying rights depending on subscription tiers. Some interpretations of the ToS suggest that commercial rights to generated images are contingent on paying membership during generation.
However, even if you “own” usage rights by contract, that does not guarantee enforceable copyright in the work itself (due to the authorship barrier).
The legal waters are shifting rapidly, and several high-profile lawsuits are now shaping precedent.
In June 2025, Disney and NBCUniversal sued Midjourney for copyright infringement, alleging that the AI generator produced images replicating their iconic characters (e.g. Yoda, Minions) without authorization. (axios.com) The plaintiffs claim Midjourney scraped their IP for training and then enabled users to recreate it. (afslaw.com) Midjourney counters that its training process is a “transformative” fair use and defends freedom of expression.
In September 2025, Warner Bros. Discovery filed a lawsuit against Midjourney, accusing the company of willful infringement of characters such as Superman, Bugs Bunny, and Wonder Woman. (ipwatchdog.com)
The U.S. Copyright Office has consistently denied copyright registration when human creative contribution is minimal. For example, in the case of Théâtre D’Opéra Spatial, the Review Board found the AI contribution dominated and declined registration. (wikipedia.org) Similarly, attempts to register comics with Midjourney illustrations have failed due to non-disclosure of AI usage. (wikipedia.org)
AI image generation stands at the crossroads of creativity and law. Tools like DALL·E and Midjourney empower users but also challenge long-standing IP principles. Courts and lawmakers are still deciding whether pure AI creations deserve protection — until then, users should act prudently, document their processes, and stay informed about the latest rulings.
Comments
Post a Comment