Retrospective 2025: Newsletter #012/2025 – Artificial Intelligence

Retrospective 2025: Newsletter (#012/2025) on Artificial Intelligence by Campos Thomaz Advogados

Alerts, materials, and updates on Artificial Intelligence.

To subscribe, click here.

In this special edition, we present a retrospective of the key materials, alerts and updates in Artificial Intelligence that shaped the year 2025. We selected the most significant regulatory, judicial and technological developments to provide a structured overview of the trends that defined the national and international landscape throughout the year. Enjoy the reading.

EU proposes postponing AI rules and easing requirements for high-risk systems

On 19 November 2025, the European Commission introduced proposals aimed at postponing key elements of the Artificial Intelligence Act and reducing regulatory burdens viewed as excessive by the technology sector. The plan includes extending the compliance deadline for high-risk AI systems – used in areas such as critical infrastructure, security, and healthcare – from August 2026 to late 2027. The package also provides for adjustments to data-protection rules, enabling broader access to datasets, including personal data, when justified by legitimate interests, with the objective of supporting the training of AI models. Learn more.

CAM-CCBC issues guidance on the use of AI in arbitration and mediation

On 17 July 2025 the CAM-CCBC issued Administrative Guidance No. 07/2025 setting out parameters for the use of artificial intelligence tools in arbitration, mediation and other appropriate dispute resolution methods, acknowledging potential efficiency benefits while cautioning against risks such as inaccuracies, algorithmic bias, compromise of confidential data and possible effects on impartiality and procedural validity. The guidance notes that the institution does not currently employ AI in its administration and that any future implementation will require additional regulation, encouraging participants to discuss the use of such tools in advance, including disclosure needs and specific procedural clauses, with decisions on AI-related matters to be made by arbitrators, mediators or dispute committees as appropriate. Learn more.

Regulatory sandbox on artificial intelligence

In June 2025 the Brazilian National Data Protection Agency launched the call for its first regulatory sandbox focused on artificial intelligence and data protection, an initiative aimed at fostering experimentation with innovative solutions centered on algorithmic transparency; applications, which were free of charge and limited to legal entities with suitable technical and economic capacity, were accepted until 10 August 2025 and are now closed. The program, scheduled to run until December 2026, will select three participants to undergo selection and development phases within an experimental environment, assessed on criteria such as innovation, project maturity, positive social impact and adherence to transparency practices, including a webinar held on 15 July and publication of the selected projects from 2 October 2025. Learn more.

ANPD Technical Note Provides Guidelines on AI and Data Protection

On 15 May 2025, the Brazilian National Data Protection Authority (Agência Nacional de Proteção de Dados – ANPD) published Technical Note No. 12/2025, consolidating the contributions received during its public call for input on artificial intelligence (AI) and personal data protection. The consultation had taken place between November 2024 and January 2025, and the Technical Note gathered diverse perspectives — including legal professionals, private-sector representatives, academia and civil society — on the regulatory challenges and possible pathways to align the General Data Protection Law (LGPD) with the growing use of automated systems, especially those based on AI. The analysis was structured around 15 questions grouped into five thematic blocks, covering LGPD principles, legal bases, data subject rights, best practices and regulatory guidelines. Learn more.

Government establishes management group for the Brazilian Artificial Intelligence Plan

On 12 May 2025 the federal government published in the Official Gazette the creation of a working group responsible for coordinating implementation of the Brazilian Artificial Intelligence Plan, which allocates R$ 23 billion and brings together 15 bodies under the leadership of the Ministry of Science, Technology and Innovation, with objectives such as developing advanced technological infrastructure, producing national high-performance processors and upgrading the supercomputer of the National Laboratory for Scientific Computing to rank among the world’s top five powered by renewable energy, underscoring AI’s strategic role in the country’s digital transformation. Learn more.

Brazil Unveils National Data Center Policy

On 5 May 2025 Brazil’s Finance Minister presented the forthcoming National Data Center Policy at the 28th Milken Institute Global Conference in Los Angeles, highlighting it as a key pillar for expanding the country’s digital infrastructure. The initiative anticipates tax-reform effects by reducing the burden on investments and service exports while prioritizing sustainable and energy-efficient data center development. Learn more.

CNJ issues rules on AI use in the Judiciary

On 18 February 2025 the National Council of Justice approved a resolution establishing parameters for the use of artificial intelligence in the Judiciary, confirming that AI may only operate as an auxiliary tool and cannot replace human decision making, while ensuring compliance with the Constitution and the General Data Protection Law. The measure, later formalized as CNJ Resolution 615/2025, sets security requirements, restricts autonomous uses, prohibits automated profiling or behavioural assessments and permits only support functions such as case organization and inconsistency detection, also creating a National Committee on Judicial Artificial Intelligence to oversee the ethical and secure adoption of these systems. Learn more.

EU AI Act: Key prohibitions take effect

As of 2 February 2025 Chapters I and II of the EU AI Act are in force, introducing rules on prohibited AI practices aimed at protecting fundamental rights. The regulation establishes a tiered risk framework ranging from minimal to unacceptable, requiring transparency for limited-risk systems, enhanced obligations for high-risk applications and banning uses that exploit human vulnerabilities or rely on biometric identification to infer sensitive traits; full implementation of the Act is expected by August 2026. Learn more.

Folha de São Paulo files lawsuit against OpenAI over unauthorized use of news content

On 20 August 2025 the newspaper Folha de São Paulo filed a lawsuit before the São Paulo state court against OpenAI, alleging unfair competition and copyright infringement arising from the use of its articles — including paywalled material — to train ChatGPT models without authorization or compensation. The complaint seeks to halt the alleged practice, obtain damages and compel the destruction of any AI models trained on the outlet’s content, citing technical records showing more than 45,000 accesses by OpenAI crawlers in July 2025 and public repository evidence suggesting data extraction from the newspaper’s domain, as well as instances in which ChatGPT reproduced full articles on the day of publication, purportedly harming digital traffic and the outlet’s subscription- and advertising-based business model. Learn more.

German court rules that ChatGPT infringed copyright by reproducing protected song lyrics, granting victory to GEMA

Germany’s collecting society GEMA secured a landmark judgment against OpenAI concerning the use of copyrighted song lyrics in training and operating ChatGPT. On November 11, 2025, the court held that the system stores and makes available copies of original works in response to user prompts, amounting to unauthorized reproduction. The court rejected OpenAI’s argument that it qualifies as a privileged research organization and clarified that text and data mining exceptions do not permit the storage or output of protected lyrics. Learn more.

High Court of Justice rules partially against Getty Images in dispute over AI image generator

On 4 November 2025, Getty Images faced a significant setback in its lawsuit before the UK High Court against Stability AI concerning the use of its image library to train the Stable Diffusion model. Getty alleged copyright and trademark infringement, arguing that its images were used to train the system and that generated outputs replicated protected content. However, midway through the proceedings, Getty withdrew the primary copyright claim due to insufficient evidence about where the model had been trained, narrowing the broader implications of the case. Learn more.

Anthropic agrees to pay USD 1.5 billion to settle authors’ class action lawsuit

On September 6, 2025, Anthropic informed a federal court in San Francisco that it had agreed to pay $1.5 billion to settle a class action brought by authors who claimed their books were unlawfully used to train the Claude chatbot. The settlement amounts to approximately $3,000 per 500,000 downloaded books and may increase if additional works are identified. The plaintiffs asked District Judge William Alsup to approve the agreement, describing it as the largest copyright recovery in history and the first of its kind in the age of artificial intelligence. Learn more.

Learn about our AI Ethics as a Service

We have prepared specific material to explain how to engage AI Ethics as a Service (external). Contact our team.

Much more about Artificial Intelligence

Discover our series of content on Artificial Intelligence. Access the full series.

Produced by Alan Campos Thomaz and João Marcelo Oliveira

share