Understanding Fair Use: The Four Factors in Copyright Law

Canadian copyright law employs a closed list of enumerated acceptable purposes for fair dealing, while US copyright law uses an open, non-exhaustive approach; additionally, Canadian law does not recognize the concept of transformativeness, which is central to US fair use doctrine.

Fair dealing is an exception to copyright infringement incorporated in the copyright statutes of common law jurisdictions such as the UK, India, and Canada. Fair use, in contrast, is a limitation to exclusive rights granted to authors under the US Copyright Act. The key difference is that fair dealing is a statutory exception while fair use is a judicial doctrine developed through case law. Both concepts serve similar purposes of balancing copyright protection with public interest in accessing and using creative works.

Most Commonwealth countries don't have fair use but have 'fair dealing' (UK, Canada). Canadian fair dealing law approximates or converges with US fair use law partly because the Supreme Court of Canada has taken a progressive view emphasizing public interest and freedom of speech. UK courts have not taken a similar line, so filmmakers can do much more under Canadian fair dealing than UK fair dealing.

Fair use is an American concept; other countries use equivalent doctrines like Canada's 'fair dealing.' Canadian fair dealing places greater emphasis on transformative nature for criticism and review. Creators using content from UK or other jurisdictions must research local laws, as they tend to be less permissive than American fair use standards.

Despite different legal frameworks, U.S. fair use and Canadian fair dealing permit similar educational uses for open educational resources, as both systems balance copyright holders' economic interests against users' rights; the key difference is that Canadian fair dealing requires uses to first fall within enumerated purposes (now including education) before applying a fairness analysis, whereas U.S. fair use immediately applies a four-factor balancing test, though both ultimately favor transformative educational uses that add new meaning or value to the original work.

Fair use should be understood as an assessment of risk rather than a binary yes/no determination. At one end of the spectrum, certain uses are clearly fair use with no doubt. At the other end, some uses are clearly infringement. Most documentary producers operate somewhere in between, and lawyers help push uses toward the safer end of this risk spectrum by applying principles that increase the likelihood of fair use protection.

A practical checklist to help determine if your use of copyrighted material is likely to be considered fair use includes: (1) Did you change the original material to make something new and different? (2) Are you using the material for non-commercial or educational purposes? (3) Did you use only a small amount of the original material? (4) Did you use the material in a way that doesn't take away from the original work's value? (5) Did you give credit to the original creator? (6) Was the original material based on facts or was it made up? (7) Was the original material already published or never released before? (8) Are you using the material in a way that doesn't hurt the market for the original work? (9) Are you using the material in a way that doesn't hurt the creator's reputation? (10) Did you ask for permission from the owner if possible? (11) Does your use benefit the public? (12) Are you using the material in a way that could be considered fair? (13) Are you using the material in a way that adds something new and useful to society? (14) Did you make sure not to use the most important or significant parts of the material? (15) Did you make sure not to use the material for something illegal?

Fair use decisions require comprehensive risk assessment rather than simple yes/no determinations. Key factors include: (1) contract limitations that may restrict fair use rights; (2) likelihood of copyright owner awareness and potential lawsuit; (3) litigation costs versus licensing expenses; (4) alternative solutions like DMCA protections or public performance exemptions; and (5) balancing institutional benefits against risks. Public benefit extends beyond transformative use to include preservation and accessibility. The optics and appearance matter significantly in court decisions. This framework recognizes fair use as part of broader risk management rather than the sole determinant.

A practical checklist for ensuring fair use compliance includes: 1) Always include an original voiceover or script providing educational or entertainment value, 2) Keep individual copyrighted clips under 10 seconds, 3) Interleave stock footage between copyrighted clips, 4) Give credit to original sources in video descriptions when appropriate. Following these guidelines helps protect against copyright claims.

Fair use analysis involves two key components: (1) evaluating the four-factor test (purpose and character of use, nature of the copyrighted work, amount and substantiality of the portion used, and effect on the market), and (2) conducting risk assessment to help filmmakers make informed decisions about what to license versus what to use under fair use. Most documentary filmmakers follow best practices guidelines and rarely face actual lawsuits, with most disputes being resolved through informal conversations and reasonable settlements.

The US has witnessed approximately 100 active copyright infringement cases against AI developers since generative AI models launched, with filings increasing dramatically from 8 in 2023 to 13 in 2024, then more than doubling to 27 in 2025. About half are class actions by groups of authors against large language model developers, though visual artists, musicians, publishers, and programmers have also sued. Two landmark summary judgments in Bard v. Anthropic and Cadri v. Meta established that training AI models on copyrighted books without permission constitutes fair use when the output isn't infringing, because this use is spectacularly transformative under fair use factor one and plaintiffs failed to demonstrate cognizable market harm. Training requires over 10 trillion tokens—data volume equivalent to the entire internet many times over—creating a market failure where no single entity could provide sufficient data. US copyright law requires human authorship for copyrightability; the monkey selfie case confirmed non-human entities cannot hold copyright. The Copyright Office has registered works using AI when there's meaningful human contribution, distinguishing between autonomous AI outputs and human-directed AI-assisted creation.

The legal framework for generative AI training on copyrighted material remains uncertain. OpenAI whistleblower Suchir Balaji argued that fair use is an implausible defense for generative AI products because they can create substitutes that compete with the data they're trained on. This raises significant legal and ethical questions about the boundaries of AI training data and the rights of content creators. The intersection of copyright law and machine learning represents a complex legal frontier that will likely require new frameworks and precedents to address.

Training AI on copyrighted material raises complex copyright questions where infringement versus fair use depends on specific facts including the purpose of use, amount of material used, and market impact; courts have ruled differently in similar cases (e.g., Google Books found fair use while Internet Archive's controlled lending was not), and fair use is not automatically available for AI training but remains a viable defense depending on the circumstances.

The US fair use doctrine considers four factors: (1) the purpose and character of the use (including whether it is transformative), (2) the nature of the copyrighted work, (3) the amount and substantiality of the portion used, and (4) the effect of the use on the potential market for the copyrighted work. Courts analyze whether AI training is transformative and whether it affects the market for original works. The Anthropic v. Authors case found that training on pirated content did not constitute fair use, while training on legitimately purchased content did. The Meta v. Authors case found that training on copyrighted works for the purpose of creating a chatbot that could compete with human authors weighed against fair use. The Perplexity v. Authors case involves allegations that Perplexity's AI system copied and reproduced news articles, presenting significant analytical challenges because it involves comparing large numbers of articles to AI-generated outputs. Copyright law is territorial, meaning it applies based on where the infringing act occurs, not where the copyright holder is located. If the infringing act occurs in Japan, Japanese copyright law applies regardless of where the copyright holder is located. Patent law and copyright law apply differently to AI training. Patent infringement requires actual implementation of the patented invention, while copyright infringement requires reproduction of protected expression. AI training typically does not constitute patent infringement because it does not implement the patented invention. However, if AI outputs implement patented inventions, patent infringement may occur.

The use of copyrighted materials to train generative AI models raises fundamental questions about fair use. Proponents argue training constitutes transformative use because AI learns patterns to create entirely new works, similar to how search engines and Google Books transformed copyrighted materials for information retrieval. Historical precedents show courts have permitted transformative functional uses even when substantial portions are used. The key question is whether training AI to generate novel creative outputs represents sufficiently transformative purpose to qualify as fair use. Courts are currently weighing this in over 30 cases nationwide, with transformative use and market effect being the most determinative factors.

This segment explains the fair use doctrine as an equitable rule that prevents rigid copyright application from stifling creativity. The court determined that reaction videos are protected under fair use only when they add additional material, commentary, and value beyond merely watching content. A reaction video that simply sits and watches content without adding substantive commentary remains vulnerable to copyright claims. The court emphasized that the fair use inquiry does not ask whether criticism is accurate or mean-spirited, but whether the use is of the kind copyright is designed to protect. Even wrong or offensive criticism can be protected by fair use.

Reaction videos often involve interspersing short segments of another's work with criticism and commentary. Courts distinguish between reaction videos with substantial commentary (more likely fair use) and group viewing sessions without critique (less likely fair use). The amount used matters less than the necessity for commentary—substantial copying is acceptable if necessary for the transformative purpose. Courts examine whether copied portions are essential to conveying the critique. The key distinction is whether the use transforms the original work or merely displays it as a prop. Critical commentary doesn't usurp market demand—it provides a fundamentally different experience than the original work.

Fair use is a legal doctrine that allows limited use of copyrighted material without permission for purposes such as criticism, commentary, or parody. The court distinguished between 'old school dumb reaction videos' where people just sit and watch without commentary, and videos with genuine commentary, editing, and transformation. The court ruled that reaction videos with substantial commentary and transformation constitute fair use, while passive viewing without meaningful engagement does not. This distinction is crucial for content creators to understand when creating reaction content.

Fair Use applies to various creative formats including: (1) Remix artists like Girl Talk and Kutiman who create new music by mashing up existing artists; (2) Gaming streams from inside copyrighted game worlds; (3) YouTube Poop - humorous video mashups created by editing pre-existing media sources; (4) Honest Trailers and Bad Lip Reading channels; (5) Supercuts and Vidding (video editing of existing footage).

Fair use is a legal doctrine that allows limited use of copyrighted material without permission from the copyright holder. The doctrine covers four main factors: (1) the purpose and character of the use (such as commentary, criticism, or reaction videos), (2) the nature of the copyrighted work, (3) the amount and substantiality of the portion used, and (4) the effect of the use on the potential market for the original work. Reaction videos that include commentary and criticism of the original content are typically considered fair use, as they transform the original work and add new meaning or message.