Authenticity and copyright infringement

  • 5 minutes

Since GenAI was trained on large datasets including copyrighted works, AI-generated content can accidentally reproduce or closely resemble copyrighted material. Nonprofits must be careful to not publish or distribute content that infringes on someone else's rights. This can be done by following a clear set of safeguards.

Use clear prompts with constraints

Avoid pasting copyrighted text into prompts.

  • Avoid asking AI to mimic a specific author/artist's style.
  • Avoid referencing known melodies, characters, or footage.
  • Ask the AI directly: "Does any part of this content resemble or match existing copyrighted text, image, multimedia?"
  • Give AI clear prompts to avoid copyrighted material.
  • Ensure original sources are cited if summarizing, explaining, or building on known materials.

Source-checking mindset and human oversight

  • Check text outputs by using plagiarism checkers (free or low-cost tools, such as those that may be part of your IT ecosystem).
  • Search key phrases in a search engine to ensure they don't exactly match existing published works.
  • Reduce the risk of using images or text without verifying rights.
  • Use reverse image search to check if the image resembles existing photographs, stock images, or illustrations.

The legal landscape surrounding AI and copyright is complex and rapidly evolving. To navigate this landscape, nonprofits must stay informed and exercise caution. However, the more pressing concern for nonprofits is the risk of violating existing copyrights.

In most jurisdictions, copyright law requires human authorship for protection, meaning that content generated solely by AI, without substantial human creative input, may not be eligible for copyright protection at all. This creates uncertainty for organizations using AI tools, as the legal ownership and protectability of AI-generated materials remain unsettled.

International copyright laws address the protection of works generated by AI. However, these laws primarily protect human authors, leading to debates over whether AI-generated content can be copyrighted and, if so, who holds those rights.

Copyrighted material

AI models are trained on vast datasets that often include copyrighted material such as articles, images, books, and creative works. AI-generated outputs can inadvertently reproduce or closely resemble protected content.

While many AI providers claim their use of copyrighted training data falls under fair use or similar legal exceptions, this interpretation is being challenged in courts worldwide, and legal precedents are still being established. For nonprofits, this means exercising caution: avoid publishing AI-generated content that too closely resembles existing works, always verify sources when AI references specific materials, and consult legal counsel when using AI for public-facing content, fundraising materials, or program documentation. As regulations continue to develop, organizations should stay informed about emerging compliance requirements in the jurisdictions where they operate.

AI tools that use copyrighted materials for training data have sparked global discussions on copyright infringement. Countries are developing different legal frameworks to address whether using copyrighted works without permission for AI training constitutes infringement.