BRAZIL Law and Practice Contributed by: Fabiano de Bem da Rocha, Gustavo Bahuschewskyj Correa, Kamille Trindade Machado and Milton Lucídio Leão Barcellos, Leão Intellectual Property
In business practice, companies typically address ownership of AI-assisted outputs through contractu- al arrangements, including employment agreements, contractor agreements and specific intellectual prop- erty assignment clauses, ensuring that any rights aris- ing from human contributions are duly assigned to the company. 7.3 Training Data, Model Development and Infringement Risk The use of copyrighted works or trade secrets in AI training datasets raises significant legal risks under Brazilian law. The reproduction of protected works for training purposes may constitute copyright infringe- ment if not properly authorised, particularly given the absence of a broad “fair use” doctrine. Brazilian law 9,610/1998 provides for specific and limited exceptions, and there is currently no clear text-and-data-mining exception comparable to those found in some other jurisdictions. The use of confi- dential information or trade secrets without authori- sation may also give rise to civil liability under unfair competition rules. Enforcement mechanisms include injunctions to cease unlawful use, damages for material and moral losses and, in certain cases, evidentiary measures such as search and seizure. Courts may also order the disclo- sure of relevant information, although this is subject to proportionality and confidentiality considerations. The legal landscape in this area is still evolving, par- ticularly as courts begin to address disputes involving AI training practices. 7.4 Enforcement Against AI-Enabled Infringement Brazilian law does not establish a specific regime for AI-enabled infringement; instead, existing intellectual property and civil liability frameworks are applied. Rights-holders may pursue claims against parties that use generative AI tools to produce infringing content, provided that the elements of infringement, includ- ing unlawful use and a causal link to the damage, are established. Courts tend to assess evidence pragmatically, con- sidering the available technical and factual elements
to determine whether the AI-generated output repro- duces protected material or creates a likelihood of confusion. Issues of causation may be more complex where AI systems are involved, but this does not pre- clude liability. Platform operators and tool providers may, in certain circumstances, be exposed to secondary liability, par- ticularly if they contribute to or fail to act upon infring- ing activities once notified. The Brazilian Civil Rights Framework for the Internet ( Marco Civil da Internet ) provides a general safe harbour regime, under which liability may depend on compliance with court orders to remove infringing content, even though recent deci- sions of the Superior Court of Justice are demanding a proactive posture of the platforms/marketplaces. As with other aspects of AI regulation, this area remains in development, with increasing scrutiny from courts and regulators. 8. Intersections and Overlaps 8.1 Choice of Protection and Strategic Considerations In Brazil, the choices among IP protection regimes are a strategic decision driven by the nature of the sub- ject matter, the desired level of exclusivity, and stra- tegic and commercial objectives. Patent protection is typically preferred when the invention can be fully disclosed without compromising its competitive value and when the applicant seeks enforceable exclusive rights for a defined term. Trade secret protection, on the other hand, is chosen when the technical information is difficult to reverse- engineer (a rarity nowadays), when patent eligibility is uncertain or when perpetual protection is commer- cially preferable over a time-limited monopoly. For software, businesses often rely on copyright com- bined with patent (when it is possible) and even trade secrets. Brand owners frequently pursue simultane- ous trade mark registration and copyright protection for logos and artistic elements to maximise enforce- ment options.
25 CHAMBERS.COM
Powered by FlippingBook