VENEZUELA Law and Practice Contributed by: Dana Bentata, Anette Beyer and Biddy Fraga Bentata, Bentata Abogados
• contract, where it violates website terms, platform rules or access conditions; and • trade secret, where non-public datasets are extracted or used without authorisation. The Artificial Intelligence Bill ( Anteproyecto de Ley de Inteligencia Artificial , the “AI Bill”), which has not yet been enacted, does not establish a dedicated scrap- ing regime or proprietary rights in data. 7.2 AI-Generated Works and Inventorship/ Authorship Venezuelan law currently contains no provisions spe- cifically addressing the authorship or inventorship of AI-assisted or AI-generated outputs. The governing IP statutes, the LPI and the LDA, both pre-date mod- ern machine learning and proceed on the basis of a human creative model, expressly defining inventors and authors as natural persons. The nation’s broader AI framework likewise does not expressly address questions of authorship or inven- torship. At present, that framework consists of the AI Bill, which was approved in first discussion before the National Assembly in November in 2024 but has not yet been enacted, and the Code of Ethics for the Responsible Development and Application of Artifi- cial Intelligence ( Código de Ética para el Desarrollo y Aplicación Responsable de la Inteligencia Artificial , the “Code of Ethics”), issued in February 2026. The AI Bill is expected to establish binding obligations once enacted, with its current focus centred on:
• privacy; • transparency; • accountability; • open science; and • excellence.
Although its accountability principle makes clear that liability for AI-driven outcomes ultimately remains with human actors, it is silent on ownership or IP entitle- ment. Even so, the Code of Ethics implies a human- centred approach to attribution. Neither instrument, however, expressly addresses attribution, inventor- ship, authorship or ownership of AI-generated out- puts. In the absence of clear statutory or administrative guidance, ownership of AI-generated outputs should be managed through carefully drafted contracts and internal policies. Where AI forms part of the creative or inventive process, parties should do the following: • clearly allocate rights in the resulting outputs; • secure appropriate assignments from employees and contractors; • ensure sufficient human involvement in the creative or inventive process to preserve the prospect of intellectual property protection; and • consider including arbitration clauses in the rel- evant agreements. 7.3 Training Data, Model Development and Infringement Risk The use of unauthorised copyright-protected material or trade secrets in model training may give rise to lia- bility under the applicable protection frameworks, but only to the extent that such use constitutes infringe- ment of those specific rights. There is no standalone statute addressing intellectual property rights in train- ing data or model outputs. The AI Bill does not confer proprietary rights in train- ing data or specifically regulate the use of protected material for model training. If enacted, it would how- ever impose standards concerning the quality, rep- resentativeness and handling of data. The penalties contemplated by the AI Bill are tied to the level of harm caused by the use of AI rather than to the unauthor- ised acquisition or use of the underlying data as such.
• consent; • privacy;
• transparency; • security; and • the responsible handling of data.
On the other hand, the Code of Ethics is a non-bind- ing instrument that sets out nine guiding principles, including: • human-centred AI;
• equity and non-discrimination; • environmental responsibility; • security;
427 CHAMBERS.COM
Powered by FlippingBook