Key Takeaways
- The European Parliament has adopted a resolution requiring generative AI providers in the EU to disclose copyrighted training data — and face potential operational bans for non-compliance.
- The US Supreme Court recently declined to hear Thaler v. Perlmutter, effectively affirming that works generated solely by AI, without meaningful human authorship, cannot be copyrighted in the United States.
- As DALL-E, Midjourney, and Stable Diffusion navigate these diverging legal frameworks, their commercial viability will increasingly depend on demonstrating human creative input and meeting transparency obligations. Two major legal developments have landed in the same week — one from Brussels, one from Washington — and together they are reshaping the rules for every business or creator using AI-generated art commercially. The EU is moving to compel disclosure and threaten operational bans. The US is holding firm on the principle that AI alone cannot be an author. For platforms like DALL-E, Midjourney, and Stable Diffusion, the implications are immediate and significant.
Generative AI Art’s Copyright Crossroads
On March 10, 2026, the European Parliament adopted a resolution on copyright and generative AI, setting a clear agenda for legislative reform. It establishes that EU copyright law applies regardless of where an AI system is trained, and introduces robust transparency requirements covering the content used in training datasets. Just days earlier, on March 2, the US Supreme Court declined to hear Thaler v. Perlmutter, affirming the long-held US position that human authorship is a foundational requirement for copyright protection. These parallel but distinct actions reflect a global struggle to define ownership and accountability in the age of algorithmic creativity — and the emerging audit and compliance requirements that come with it.
DALL-E: OpenAI’s Structured Ecosystem
OpenAI’s DALL-E operates within a relatively controlled commercial environment. Its terms of service and content policies are designed to provide clearer guidance for businesses and individual creators navigating generative AI, and OpenAI generally permits commercial use of DALL-E outputs under defined conditions.
The platform uses a structured API with built-in content safeguards aimed at reducing the generation of problematic or infringing material. Users are generally treated as owners of the images they create, though the broader legal enforceability of copyright for purely AI-generated elements remains unsettled. OpenAI’s explicit contractual language on output ownership gives users a working framework — but that framework still sits within the wider legal uncertainty surrounding AI-generated content. The model’s training on large datasets, including potentially copyrighted material, remains part of the industry-wide debate over fair use and infringement.
Midjourney: Artistic Experimentation Meets Legal Challenges
Midjourney has built a reputation for sophisticated image generation and creative experimentation, largely accessed through a Discord interface. That artistic freedom, however, has come alongside some of the most significant copyright litigation facing any AI platform. Artists have alleged that Midjourney, along with other AI companies, infringed on their rights by training on large volumes of images scraped from the web without consent.
Commercial rights on the platform are tiered — paid subscriptions are required for commercial use, with higher-tier plans required for companies above certain revenue thresholds. Users generally retain ownership of their generated images, though Midjourney keeps a licence to use and display them for service promotion. Ongoing litigation, including cases alleging copyright infringement for the use of protected characters and artworks in training data, means the platform faces sustained legal scrutiny that users operating commercially cannot afford to ignore.
Stable Diffusion: The Open-Source Frontier
Stable Diffusion’s open-source model gives users the greatest flexibility — including the ability to run it locally and customise outputs extensively. That freedom, however, carries the highest individual legal exposure. Unlike DALL-E or Midjourney, Stable Diffusion’s developer provides no indemnification to end users for copyright issues that may arise from their use of the model.
The training data underpinning Stable Diffusion — notably the LAION datasets compiled from images scraped across the internet — has been central to several high-profile lawsuits. Getty Images v. Stability AI alleges the unauthorised use of large volumes of copyrighted images, while Andersen v. Stability AI involves artists claiming direct copyright infringement. The outcomes of these cases are likely to have meaningful consequences for the broader Stable Diffusion ecosystem, even if direct end-user liability currently appears limited. Despite the legal uncertainty, the open model continues to drive rapid innovation and remains the platform of choice for developers and creators who want maximum control over their process.
The Copyright Crossroads: EU’s Transparency Mandate vs. US’s Human Authorship Rule
The divergence between US and EU approaches is becoming the defining tension in AI copyright policy. In the United States, human authorship remains the threshold requirement. The US Copyright Office has consistently held that works generated entirely by AI, without meaningful human creative input, are ineligible for copyright and fall into the public domain by default. The Supreme Court’s decision in Thaler v. Perlmutter closed off the most direct challenge to that position, ending a sustained effort to secure copyright for an autonomously generated work. The Copyright Office does, however, permit registration for AI-assisted works, provided the human author discloses the AI’s role and demonstrates sufficient creative contribution — through prompt engineering, selection, arrangement, or post-generation editing.
The EU is taking a markedly different approach. The European Parliament’s resolution mandates that generative AI providers operating in the EU comply with EU copyright law regardless of where their models were trained. Critically, it requires providers to supply an itemised list of each piece of copyright-protected content used in training — a transparency obligation with no equivalent in US law. The resolution goes further, indicating that providers failing to comply should be barred from operating within the Union. This represents a deliberate effort to prevent companies from bypassing EU standards by training models in lower-compliance jurisdictions. These measures sit alongside existing provisions in the EU AI Act, which separately governs transparency and data documentation requirements for high-risk AI systems.
Which Should You Choose? Navigating the Legal Maze
For commercial and professional users, the choice between these platforms is now as much a legal and compliance decision as a creative one.
Risk-averse businesses seeking clear terms and a managed environment will find DALL-E the most structured option. OpenAI’s explicit commercial use policies provide working clarity on ownership, even if the broader legal questions around AI-generated content remain open.
Artists focused on stylistic range and rapid iteration may still favour Midjourney, but its ongoing litigation and tiered commercial licensing mean users need to stay across both their subscription terms and any shifts in case law. Documenting human creative input carefully is increasingly important.
Developers and creators who need maximum customisation will gravitate toward Stable Diffusion, but should do so with a clear understanding that the open-source model places the full weight of copyright compliance on the individual user. The outcomes of current litigation could change that calculus significantly.
Across all three platforms, the practical advice is consistent: document your creative process thoroughly, make substantive human contributions at the prompt, editing, or curation stage, and disclose AI use where required. As jurisdictions continue to develop competing frameworks, staying informed about both regional law and platform-specific terms is no longer optional for anyone using these tools professionally. For more coverage of AI policy and regulation, visit our AI Policy & Regulation section.
Originally published at https://autonainews.com/dall-e-midjourney-stable-diffusion/
Top comments (0)