The rapid spread of generative artificial intelligence (AI) — on platforms such as ChatGPT, Midjourney, or DALL·E — demands an examination of centuries-old principles regarding authorship, creation, and intellectual property protection in the digital environment. The outputs derived from these systems challenge the distinction between human creation and results obtained through autonomous algorithms.

Therefore, the classical theory of legal fact — developed within the traditions of civil law — presents itself as a promising interpretative lens for determining when a production mediated by AI can acquire the status of a protected work or when it should remain at the level of a mere fact.

In general legal theory, legal phenomena are traditionally organized into three stages: existence, validity, and effectiveness.

In the context of copyright, it is not enough for something to simply "exist." It requires repercussions in a person's legal sphere, and in this case, creative human authorship, to attract legal protection. In other words, the mere existence of an AI-generated file does not seem to guarantee that it will automatically be treated as a protected work; relevant human intervention is mandatory.

In the Brazilian legal system, Article 11 of Law 9.610/1998 states that "Author is the natural person who creates a literary, artistic or scientific work".¹

This has been interpreted as indicating that, as a rule, natural persons are recognized as authors, and that works entirely produced by AI, without significant human contribution, do not satisfy this authorship requirement. However, it is worth remembering the possibility that a legal entity may be the derivative holder of patrimonial rights (by assignment).

Law 9.609/1998 deals with the protection of intellectual property rights for computer programs, defined as "the expression of an organized set of instructions in natural or coded language, contained in physical media of any nature, necessary for use in automatic information processing machines, devices, instruments or peripheral equipment, based on digital or analog technology, to make them function in a determined manner and for specific purposes." In other words, it does not seem to automatically extend to outputs produced by AI, except when such outputs are integrated into a work or contract involving relevant human creative intervention, leaving the discussion open precisely regarding prompts.

Thus, the Brazilian authorship framework reveals significant gaps when dealing with autonomous or semi-autonomous creations resulting from AI systems.

In the international arena, in the United States, there is the US Copyright Office (USCO), which published the report. Copyright and Artificial Intelligence: Part 2 – Copyrightability em 29 January 2025.³

In this report, the USCO reaffirms that human authorship remains central to copyright protection, and that works whose expressive elements are determined exclusively by AI without creative human intervention are not registrable.⁴

Additionally, the report highlights that human selection, editing, or creative arrangement of AI outputs can generate protectable works, provided they surpass the minimum threshold of originality.⁵

In the European Union, Regulation (EU) 2024/1689 – AI Act establishes a horizontal set of rules for AI systems, with a risk-based approach (categories: unacceptable, high, limited, minimal) and obligations of transparency, traceability, auditability, and accountability.⁷

It is important to emphasize that the AI ​​Act does not replace EU or Member State copyright rules, but rather functions as a supplementary regulatory regime that interacts with copyright obligations, contracts, and compliance.

In the United Kingdom, the High Court delivered its ruling on November 4, 2025, in the case. Getty Images v. Stability AIThe court rejected the core of the copyright claims and recognized only limited trademark infringement for the use of the... watermark 'Getty'. The debate over training with protected material remains open in other jurisdictions (there is a lawsuit in the US).

In this regard, some questions arise: 

Is training the same as reproduction? In the case Infopaq International A/S v. Danske Dagblades Forening (C-5/08), the Court of Justice of the European Union recognizes that minimal excerpts may constitute reproduction in part if they reflect intellectual creation; the analysis is usually accompanied by verification of the exception of 'temporary copies' of art. 5(1) of Directive 2001/29.

What level of human creativity is required? According to the USCO report, a perceptible human contribution is necessary; that is, simple generic prompts hardly reach the required threshold of originality.

Transparency, governance, and algorithmic accountability? Regulatory regimes such as the AI ​​Act impose obligations regarding dataset documentation, traceability of training processes, explainability of decisions, audits, and risk reporting, requiring adaptations to compliance, contracts, and governance in organizations that use AI.

Thus, the most plausible regime for the AI ​​environment points to a hybrid model: parts of the result that incorporate human expression should be protected via copyright; purely automated components should be governed by contracts, licenses, ethical clauses, and AI-specific regulatory regimes. The great challenge lies in defining objective and transparent criteria for human creativity, algorithmic auditing, and effective accountability.

Sources:

1. BRAZIL. Law No. 9.610, of February 19, 1998. Art. 11.
2. See PONTES DE MIRANDA, Francisco Cavalcanti. Private Law TreatySão Paulo: RT. See also MELLO, Marcos Bernardes de. Theory of the Legal FactSão Paulo: Saraiva.
3. UNITED STATES COPYRIGHT OFFICE. Copyright and Artificial Intelligence: Part 2 – CopyrightabilityWashington, DC, Jan. 29, 2025. Available at: https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf. 
4. Id. Section II.
5. See additional analysis at: SKADDEN. Copyright Office Publishes Report on Copyrightability of AI-Generated Materials. February 4, 2025.
6. UK - Getty Images v Stability AI (High Court, November 04, 2025)Sentence and analysis: https://www.judiciary.uk/wp-content/uploads/2025/11/Getty-Images-v-Stability-AI.pdf?utm  
7. https://www.judiciary.uk/wp-content/uploads/2025/11/Getty-Images-v-Stability-AI.pdf
8. CJEU – Infopaq (C-5/08): ruling: https://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=CELEX%3A62008CJ0005%3AEN%3AHTML&utm
9. EUROPEAN UNION. Regulation (EU) 2024/1689 of 12 July 2024. Artificial Intelligence Act. Official Journal L 305/1, 12.07.2024.

See also: European Commission. Regulatory Framework for Artificial Intelligence. Available at: https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai.