Start 50 Plan: From R$69 to R$59/month and Plus 50 Plan: From R$119 to R$99/month.

The use of generative AI for software development: an analysis in the light of European legislation

Publicado em:

2024

Artificial intelligence (AI) has become a powerful and disruptive tool in various sectors of society, including the development of softwares. Generative AI is a type of artificial intelligence that has the potential to revolutionize the way in which softwares are created because it can automate complex tasks, optimize processes and generate innovative solutions.

In view of the fact that it is a new technology, there is a concern to adapt it to the legislation related to the subject and to ethical and safety principles.

This article aims to analyze the use of Generative AI for the development of softwares from the perspective of European legislation. The main concepts on the subject, the applications of Generative AI in this context and the challenges and opportunities that European legislation presents for the development and implementation of this technology will be addressed.‍

Basic Concepts

Generative AI

Generative AI is a promising species of artificial intelligence with transformative power for the way content is created and reproduced. It is characterized by the ability to generate new content, such as texts, images, music and codes, going beyond the previously known limits of the mere imitation used by traditional artificial intelligence, which only compiles and analyzes information.

Rather than simply processing data, Generative AI learns from large volumes of information, patterns, and complex relationships. Through machine learning techniques, especially artificial neural networks, systems learn through comprehensive data sets, internalizing patterns and rules identified by it, and are capable of creating something new. It means that, instead of following pre-established instructions, Generative AI explores the database autonomously, seeking to identify correlations, sequences, and patterns.

Examples of new content created by Generative AI are:

Intuitive and Personalized Interfaces: Generative AI can design for each user a specific interface that anticipates their needs and optimizes their experience.
Automatic Generation of Complex Codes: This can be accomplished from simple natural language descriptions, accelerating the development of softwares.
Creation of Innovative Algorithms: They can be used to solve problems before they even arise, optimizing system performance.
Songwriting: Generative AI is capable of creating innovative melodies, harmonies, and rhythms.
Works of Art: Generative AI is capable of generating new images, from abstract landscapes to personalized portraits, combining elements from different existing images. It can also modify existing images, such as changing colors, textures, adding elements, or removing objects.
Creative texts: Generative AI is capable of writing poems, screenplays, and articles from different styles and perspectives.

Companies like Google and Amazon are already using Generative AI to develop new products and services, from Chatbots intelligent to personalized translation tools. Startups innovative solutions are emerging with solutions based on Generative AI to automate repetitive tasks, streamline processes and create new experiences for users.

However, it should be noted that the originality of Generative AI is still under debate. There is a current of experts who claim that Generative AI is not yet capable of creating something totally new, only recombining existing elements in an innovative way. Another current believes that AI is evolving rapidly and that it already reaches a level of originality comparable to that of humans, since human beings also use their previous knowledge of existing elements to create new content.

Regardless of perspective, Generative AI is having a significant impact on the way we create and consume content. It's a powerful tool that can be used to broaden our creativity, explore new ideas, and develop innovative solutions. The key to success, as we will discuss later, is its responsible use, always complying with ethical principles, transparency and security.

Machine Learning

Machine learning is a type of artificial intelligence that gives computer systems the ability to learn and improve their skills without having been previously programmed. This learning occurs through data analysis, without the need for pre-defined instructions.

In this analysis, the systems identify patterns, rules, and have insights, improving your performance on specific tasks autonomously. This tool is capable of boosting progress in various sectors, including the development of softwares.

Artificial Neural Networks

Artificial neural networks (RNAs) are one of the fundamental pillars of artificial intelligence and are inspired by the structure and functioning of the human brain to solve complex problems and learn with data autonomously. They are computer systems composed of interconnected units called artificial neurons, which process and transmit information in a similar way to neurons in the human brain.

Through an interactive learning process, these networks adjust their weights and connections, becoming increasingly adept at carrying out specific tasks.

Application of Generative AI in Software Development

Automation of Repetitive Tasks

In the dynamic world of development of softwares, the search for efficiency and productivity is constant. It is in this context that the automation of repetitive tasks stands out as a powerful tool, freeing up time for more strategic and creative activities and boosting innovation and project success.

The automation of repetitive tasks in the development of softwares consists of the use of tools and scripts to automate manual, standardized tasks that consume unnecessary developer time and effort, such as code generation, version management, test implementation, and application deployment.

The implementation of the automation of repetitive tasks in the development of softwares provides benefits such as increased productivity, improved code quality, speeding up development time, greater standardization and consistency, and cost reduction.

Creating Intuitive Interfaces

In the universe of the development of softwares, the creation of intuitive interfaces stands out as an important element for successful applications. A well-designed interface, which anticipates the user's needs and guides them in a natural way, is able to transform your experience, making it more enjoyable, efficient, and productive.

The intuitive interfaces in the development of softwares are characterized by ease of use and understanding, allowing users to interact with the application in a natural and intuitive way, without the need for extensive learning or complex manuals.

Invest in the creation of intuitive interfaces in the development of softwares generates several benefits, such as improved user experience, increased productivity, reduced support costs, improved brand image, and increased customer loyalty.

Performance Optimization

In the context of Generative AI for the development of softwares, performance optimization is an important aspect to guarantee the viability and scalability of the solutions created. Through appropriate techniques and strategies, it is possible to improve the efficiency of softwares, making them faster, more accurate and robust, which boosts the productivity and quality of the final application.

Generation of Innovative Solutions

Generative AI can generate innovative solutions. Through its ability to learn with data, identify patterns, and create new information, Generative AI allows developers to push the boundaries of traditional programming, directing new paths and achieving results that could previously be unimaginable.

In the context of the development of softwares, this may mean development of Chatbots and virtual assistants, creation of personalized content, development of design, code generation and documentation, and game and simulation development.‍

European Legislation and its Impacts

The European Union's AI Act, also known as AI Act, introduces a comprehensive regulatory framework for the use of artificial intelligence in various areas, including the development of softwares. Generative AI, which, as already explained, is a promising type of artificial intelligence capable of creating new content and ideas, is subject to this law, which requires developers to adopt measures to guarantee the safety, reliability and responsibility of the use of this technology.

The AI Act classifies artificial intelligence systems into four risk categories based on their potential to cause harm:

Unacceptable Risk: Systems prohibited from being developed, including those that manipulate human emotions, exploiting emotional or psychological vulnerabilities to influence people's behavior; or score individuals' social reliability, such as social credit scores (Chapter II, art. 5).
High Risk: Systems subject to strict requirements, such as compliance assessment by an independent body and the implementation of measures to mitigate this risk. Examples of high-risk artificial intelligence are systems that manage critical infrastructures, such as energy or transportation networks; perform medical diagnoses; and make legal or law enforcement decisions (Chapter III, art. 6)
Limited Risk: Systems subject to transparency requirements, such as the obligation to inform users about the use of artificial intelligence and the provision of technical documentation (Chapter IV, art. 50).
Minimum Risk: Systems that do not have specific requirements, but must follow ethical and good faith principles, such as avoiding algorithmic biases, protecting data privacy, and ensuring user safety.

Most Generative AI systems are classified as “Limited Risk” and must meet the following requirements (Chapter IV, art. 50):

Transparency: The need to disclose that the content was generated by artificial intelligence, providing information about the models used and the training data.
Copyright Compliance: Mandatory to ensure that copyrights are respected, especially when using protected data to train artificial intelligence models.
Implementation of Security Measures: Adoption of measures to protect data and prevent artificial intelligence from being used for malicious purposes.

However, not all Generative AI applications are classified as “Limited Risk”. Several factors can influence the actual risk level of Generative AI, such as:

Type of Generated Content: The generation of realistic images or of Deepfakes, for example, presents greater risks than that of simple texts.
The Context of Use: The use of Generative AI in sensitive areas, such as health or law, requires greater care in risk assessment.
The Quality and Origin of Training Data: Biased or discriminatory data can lead to biased and harmful results.

In this context, we can interpret that the AI Law is applied to Generative AI used in the development of softwares with regard to the following principles:

Transparency and Explanability: Developers must be transparent about using Generative AI in their softwares, informing users of its functionalities, limitations, and potential impacts. Depending on software, Generative AI systems must be able to explain how they arrived at a certain result, especially when decisions that impact people's lives are made. This requires developers to issue clear and concise warnings to users, provide information about the data used, give the option to deactivate Generative AI, implement model interpretation techniques explaining how Generative AI made decisions and what factors influenced the results, and provide counterfactual explanations, demonstrating to users that the final result could have been different if other data had been provided to the system.
Legality, Impartiality, and Non-Discrimination: Generative AI systems must operate in compliance with the law, avoid algorithmic biases, and do not discriminate against individuals or groups.
Safety and Robustness: Generative AI systems must be robust and secure, protected against cyberattacks and failures that may lead to harmful or misleading results. This requires developers to improve their techniques for training and operating the models, in addition to implementing more stringent security measures.
Risk Management: Developers must implement measures to identify, assess, and mitigate risks associated with the use of Generative AI, such as algorithmic biases, discrimination, and misinformation.
Data and Training: The data used to train Generative AI models must be of high quality, free of bias, and collected in an ethical and responsible manner.
Data Protection and Privacy: The data used to train and operate Generative AI models must be collected and processed in accordance with data protection laws, such as the GDPR. This requires developers to implement measures to ensure the security and privacy of the data used, obtain appropriate consent from users, and allow them to access their data.
Governance and Oversight: Developers must implement governance and oversight mechanisms to ensure the responsible and ethical use of Generative AI in their softwares with the establishment of ethical policies.
Safety and Human Protection: Generative AI systems must be designed to ensure the safety and security of people, avoiding causing physical or psychological harm.
Responsibility: Developers must be responsible for the damage caused by their Generative AI systems, implementing mechanisms to identify, prevent, and mitigate risks.

Based on the analysis of the European Union AI Act, the following recommendations can be made for Generative AI developers:

Perform a risk assessment: Identify and assess the potential risks associated with the use of Generative AI in your softwares, considering factors such as the type of data used, the functionalities of software and the potential impact on users.
Implement risk mitigation measures: Based on the result of the risk assessment, implement measures to mitigate the identified risks, such as training models on high-quality data, implementing security mechanisms, and developing clear documentation on the functioning of Generative AI.
Test and monitor the software: Test the software to identify and correct flaws that may lead to harmful or misleading results. Monitor the software in production to detect and solve problems that may arise.
Document the development process: Document the development process of software, including decisions made regarding Generative AI, the risks identified, and the mitigation measures implemented.
Provide clear information to users: Provide clear information to users about the use of Generative AI in their softwares, including its functionalities, limitations, and potential impacts.

The European Union's AI Act (AI Act) presents the following challenges for the development of softwares with Generative AI, requiring adaptations and innovations to meet regulatory requirements.

Complexity of the Law: The interpretation of risk classification criteria and compliance with the AI Act can be complex and subjective, leading to legal uncertainty for developers.
Lack of Standards and Guidelines: There is a lack of clarity in some aspects of the AI Act, such as a more precise definition of the explanation of how the results were achieved and of the specific methods for evaluating and mitigating risks, which makes it difficult to implement appropriate measures.
Implementation Costs: The implementation of compliance measures with the European Union's AI Act (AI Act) may generate additional costs for developers, such as investments in audits, documentation, and employee training (Chapter III, Section II, Chapter 3).
Deceleration of Innovation: The time and resources needed to ensure compliance with the AI Act may slow down the pace of innovation in the area of Generative AI, which also affects more specifically startups and small businesses.
Market Fragmentation: The AI Act, as a general and comprehensive legal instrument, grants Member States of the European Union flexibility to implement specific national rules and regulations for the use of Generative AI in their territories. This flexibility, while aimed at meeting the particularities of each country, can lead to the fragmentation of the single European market, with different requirements and recommendations for developers operating in different countries. Examples of fragmentation may be variable definitions, as the AI Act does not clearly define important concepts such as “minimum risk” or “explicability”, allowing for a different interpretation between Member States; diverging compliance requirements, since each Member State can define its own documentation, testing and auditing requirements; and specific prohibition rules that may differ between Member States of the European Union.
Change in the Focus of Development: The focus of the development of softwares with Generative AI can be redirected to areas with lower regulatory risk.

In contrast, the AI Act offers the following advantages for the development of softwares with Generative AI.

Increased User Confidence: Compliance with the AI Act can increase user confidence in relation to softwares that use Generative AI, driving the adoption of the use of this technology.
Market Differentiation: Developers who demonstrate commitment to the ethical and legal principles of AI can stand out in the market and attract clients who value responsibility in the use of technology.
Promoting Responsible Innovation: The AI Act can stimulate the development of new Generative AI technologies that are safer, more transparent, and beneficial to society.
Reducing the Risk of Litigation: Compliance with the AI Act can reduce the risk of litigation related to the use of Generative AI, protecting developers, whether individuals or legal entities, from fines and other damages.
Creating a More Predictable Regulatory Environment: In the long term, the AI Act may contribute to the creation of a more predictable regulatory environment for Generative AI, which facilitates the planning and investment of investors in this area.

It is important to clarify that the European Union's AI Act (AI Act) does not apply directly to Brazil. However, the principles and guidelines of the AI Act can serve as a valuable reference for the responsible development of softwares with Generative AI in Brazil, even before the implementation of a comprehensive national law, as we will see in the next topic.

Influence of European Law in Brazil

Europe's experience with the European Union's AI Law (AI Act) offers several insights for Brazilian developers who want to create softwares with Generative AI in a responsible and ethical way, such as:

Compliance: Compliance with relevant laws and regulations that already exist in Brazil on the subject is essential for the safe development of Generative AI systems in the country, as well as to avoid legal sanctions, such as a fine for non-compliance with the law.
Ethical Principles: Generative AI must be developed and used in a way that respects human rights, data privacy, and ethical values.
Transparency and Explanability: Generative AI systems must be transparent and explainable, allowing users to understand how they work and make decisions.
Risk Management: It is essential to identify, assess, and mitigate potential risks associated with the use of Generative AI, such as algorithmic biases, discrimination, or misinformation.
Security and Data Protection: The data used to train and operate Generative AI models must be protected against unauthorized access, violations, and misuse.

Although the European Union's AI Act (AI Act) is not directly applicable in Brazil, there are measures that Brazilian developers can take to adapt their systems to the local scenario.

Monitor Regulatory Development: Monitor the development of laws and regulations related to artificial intelligence in Brazil and adapt to new requirements as they arise.
Engagement with Stakeholders: Dialogue with different stakeholders (Government, academia, civil society and private sector) to contribute to the development of a balanced national regulatory framework that contributes to innovation.
Adopt International Reference Practices: Implement international best practices in terms of the development, implementation, and use of artificial intelligence.
Search for Certifications and Quality Seals: Obtain certifications and quality seals that demonstrate the developer's commitment to the responsible and ethical development of artificial intelligence.
Promote Education and Awareness: Invest in artificial intelligence education and awareness initiatives for the general public, contributing to an informed and engaged public debate about the challenges and opportunities of technology.

The development of softwares with Generative AI in Brazil requires a commitment to responsibility, ethics, and compliance with relevant laws and regulations. By learning from European experience, adopting international best practices, and adapting to the local scenario, Brazilian developers can contribute to the advancement of this technology in a safe, advantageous and inclusive way for society.

Conclusion

Generative AI has the potential to revolutionize the development of softwares, bringing significant advantages in terms of productivity, efficiency, innovation, and user experience. However, it is essential that the development and implementation of this technology comply with current legislation and that challenges related to ethics, security, and intellectual property are adequately addressed.

Developers who intend to use Generative AI in the development of softwares should prepare for regulatory and ethical challenges by investing in training, robust documentation, and risk mitigation measures. By doing so, they can take advantage of the opportunities that this technology offers to create innovative solutions and enhance the user experience.

‍