Spain’s Council of Ministers has approved a draft bill proposing substantial fines of up to €35 million for artificial intelligence (AI) companies that fail to accurately label AI-generated content, particularly targeting the spread of misleading “deepfakes.”
Under this draft legislation, incorrectly labelled AI-generated materials are classified as a serious violation, attracting penalties ranging from €7.5 million to €35 million, or between 2% and 7% of a company’s global annual revenue. The proposed law does, however, suggest that penalties could be less severe for startups and medium-sized enterprises.
Oscar Lopez, Spain’s Minister for Digital Transformation, highlighted the law’s intent, stating via Instagram: “Artificial intelligence must help us improve the world that we live in.” Lopez emphasized that the legislation aims to ensure AI contributes positively to society – such as supporting medical research, disaster prevention, and urban traffic management – rather than facilitating the dissemination of “hoaxes, lies and defamation.”
This proposed Spanish legislation aligns with the principles established by the European Union’s AI Act, which enforces stringent transparency standards on AI systems categorized as high-risk, according to Lopez.
Before becoming law, the bill must still receive formal approval from Spain’s Congress of Deputies.
Additional Prohibitions Under the Proposed Law
The draft law additionally seeks to prohibit various other practices deemed harmful. Among these is the use of subliminal techniques, specifically methods employing “imperceptible images or sounds” designed to influence decisions without the individual’s explicit consent. For instance, such practices could involve an AI chatbot identifying individuals with gambling addictions and subsequently encouraging them to access gambling platforms.
Further restrictions included in the bill target the use of AI systems for classifying individuals based on sensitive attributes, such as race, political ideology, religious beliefs, or sexual orientation, derived from biometric data or social media activity. The proposed law explicitly bans using such data to predict criminal tendencies, responding to criticisms regarding an algorithm previously deployed in Catalonia’s prison system to assess the risk of recidivism among inmates.
Additionally, the law would penalize AI companies employing biometric systems for employee attendance monitoring without appropriate human oversight, with fines ranging from €500,000 to €7.5 million, or between 1% and 2% of their global annual turnover.
Finally, the bill grants the Spanish government authority to temporarily prohibit the operation of any AI system within the country “when they have caused a serious accident, like the death of a person.”
