Escrever um comentário

ETHICAL AUDIT OF ARTIFICIAL INTELLIGENCE ALGORITHMS: PROBLEMS AND CHALLENGES FOR MODERN LEGISLATION

Diana Radchenko 10th Grade Student State Budgetary Educational Institution Dostoevsky School, Building No. 10

Moscow, Russia

Abstract: The article addresses the problem of the lack of comprehensive legal regulation of the ethical audit of artificial intelligence algorithms. The purpose of the study is to identify the key challenges associated with the implementation of ethical standards into legislation. The research applies methods of comparative analysis of international regulatory documents, analysis of digital ethics concepts, as well as generalization of the experience of leading countries in the field of artificial intelligence regulation. As a result, gaps in the current legislation are revealed, the necessity of institutionalizing the practice of mandatory ethical audit is demonstrated, and possible ways of integrating ethical criteria into national and international legal systems are proposed. The practical significance lies in the development of proposals for creating a certification system for artificial intelligence algorithms that takes into account the requirements of social justice, transparency, and the protection of human rights.

Key Findings

The research showed that current legislation does not fully address the ethical challenges created by artificial intelligence. Many AI systems still lack transparency, making it difficult to understand how important decisions are made. The research also found that ethical auditing can help identify bias, improve fairness, and increase public trust in AI technologies. The analysis of international experience demonstrates that more countries are still introducing ethical principles into AI regulation, making this issue increasingly important.

Recommendations

Based on the results of this research, ethical auditing should become a regular part of the development and use of high-risk AI systems. New legal requirements should encourage greater transparency, explainability, and accountability of algorithms. It is also important to strengthen cooperation between governments, researchers, and technology companies to develop clear and effective rules for the responsible use of artificial intelligence.

Conclusion

This article confirms that ethical auditing can play an important role in making artificial intelligence safer and more trustworthy. As AI becomes more widely used in different areas of life, legislation should develop alongside new technologies. Creating clear ethical and legal standards will help reduce risks, protect people's rights, and support the responsible development of artificial intelligence.

ETHICAL AUDIT OF ARTIFICIAL INTELLIGENCE ALGORITHMS: PROBLEMS AND CHALLENGES FOR MODERN LEGISLATION

Introduction

Artificial intelligence is being actively integrated into healthcare, education, transportation, industry, and everyday life. While its use offers significant benefits, it also creates new challenges. Alongside technical issues, ethical concerns related to algorithm transparency, the protection of human rights, and technology regulation have become increasingly important.

One of the key mechanisms for oversight is the ethical audit of artificial intelligence systems. An ethical audit evaluates AI systems in terms of fairness, non-discrimination, and transparency. Unlike technical assessments, ethical audits focus on the impact of technologies on society and individuals, making it possible to identify hidden risks even before large-scale implementation.

This approach is necessary because many AI systems operate as “black boxes.” Even experts are not always able to explain why a system arrives at a particular decision. This lack of explainability reduces trust, complicates the process of challenging decisions, and creates risks for businesses, governments, and citizens. Therefore, transparency and explainability have become essential prerequisites for the sustainable development of artificial intelligence technologies.

If moral and legal principles are ignored, the development of artificial intelligence may cause more harm than benefit. Errors in medical algorithms, discrimination in recruitment systems, or the use of AI technologies for manipulation can undermine public trust in innovation. For this reason, the implementation of ethical auditing practices and their incorporation into legislation have become necessary conditions for the safe development of a digital society.

1. Artificial Intelligence and Its Impact on Society

Artificial intelligence (AI) is becoming one of the key drivers of modern social development. Its implementation affects the economy, social relations, education, healthcare, and public administration. Unlike previous technological revolutions, AI influences not only production processes but also human intellectual activity. Data analysis, decision-making, and the distribution of responsibility are increasingly being delegated to algorithms.

The first area of impact is the economy. AI systems automate tasks that require the processing of large volumes of data and complete them in milliseconds or seconds. Companies use artificial intelligence to forecast demand, optimize logistics, and develop new products. Through accurate planning and the automation of routine operations, organizations are able to reduce operational costs and increase efficiency. At the same time, there is a growing risk of job displacement in traditional occupations characterized by repetitive tasks. Conversely, positions involving creativity, analytical thinking, and management continue to require human participation. As a result, society faces the challenge of large-scale workforce retraining and the development of new competencies, ranging from basic data literacy to advanced skills in working with AI-based tools.

In healthcare, artificial intelligence is already demonstrating significant benefits. Computer vision technologies assist in the analysis of MRI, CT, and X-ray images, while machine learning algorithms can predict disease progression and recommend treatment options. Such systems are increasingly integrated with electronic medical records, enabling physicians to receive diagnostic information more quickly. This improves diagnostic accuracy and reduces decision-making time from many minutes down to just a few.

At the same time, the principle of human-in-the-loop remains essential. The final decision continues to belong to the physician, who may approve or reject the recommendations generated by an algorithm. Nevertheless, new questions arise: who should be held responsible for errors—the developer, the physician, or the healthcare institution? Furthermore, how should modifications to data and model versions be documented and monitored?

AI implementation also raises social risks, including discrimination and violations of personal privacy. Algorithms may reproduce existing biases if they are trained on incomplete or distorted datasets. In addition, facial recognition technologies and behavioral analysis systems directly affect individuals’ right to privacy. Therefore, effective oversight mechanisms are required, including algorithmic transparency, independent evaluations, and robust personal data protection measures.

The ethical implications of AI have become a major subject of academic and regulatory discussion. The use of artificial intelligence requires a legal framework capable of balancing innovation with the protection of human rights. International initiatives promoting the concept of “responsible AI” emphasize principles such as fairness, safety, transparency, and accountability [4; 5; 6; 9].

Thus, artificial intelligence has a complex and multifaceted impact on society. It creates opportunities for economic growth and improvements in quality of life, while simultaneously generating new risks. The future of AI integration depends on how effectively governments, businesses, and the scientific community can combine technological progress with ethical and legal safeguards.

2. Ethical Risks of Artificial Intelligence

The use of artificial intelligence (AI) raises a number of ethical challenges with significant implications for society, human rights, and legal practice. These risks should be considered in the context of protecting human dignity, safeguarding fundamental rights, and ensuring sustainable development.

Several real-world examples clearly demonstrate why the absence of ethical auditing may lead to social, legal, and economic problems.

Amazon's recruitment algorithm is one of the most well-known cases. The company tested an internal AI system designed to evaluate job applicants. The algorithm was trained using historical resumes submitted to the company, most of which belonged to male candidates. As a result, the system systematically downgraded resumes submitted by women, particularly those containing terms such as "women's club" or references to women's colleges. Ultimately, Amazon discontinued the project, but the company still faced reputational damage. An ethical audit could have identified the bias before deployment.

Another example is the COMPAS system used in the United States to predict the likelihood of criminal reoffending. Investigative journalists found that the algorithm more frequently assigned high-risk scores to African American defendants while underestimating risks for white defendants. Judges relied on these predictions when handing down sentences. Independent auditing of the system's data and algorithms could have revealed discriminatory patterns and limited the use of such flawed assessments.

Research conducted by Joy Buolamwini and Timnit Gebru in 2018 demonstrated that facial recognition systems developed by major technology companies produced significantly higher error rates when identifying women with darker skin tones. While accuracy for white males exceeded 99%, error rates for darker-skinned women reached up to 34%. Ethical auditing would require companies to test AI systems on diverse demographic groups before commercial deployment.

Another important example comes from healthcare. A study published in the journal Science revealed that a widely used medical algorithm underestimated the severity of illnesses among African American patients because it relied heavily on historical healthcare expenditures. Since healthcare spending had historically been higher among white patients, the system incorrectly concluded that Black patients required less medical attention. As a result, millions of patients may have received insufficient care. Ethical auditing could have identified this bias and required modifications to the evaluation criteria.

Why do these challenges arise?

One of the primary reasons is the lack of transparency. Many advanced AI models operate as so-called "black boxes." Users and even developers often cannot fully explain how a particular decision was reached. This lack of transparency makes it difficult to evaluate the reliability of recommendations and increases the likelihood of hidden errors. Therefore, transparency and explainability are increasingly recognized as essential requirements for trustworthy AI.

Another major issue is algorithmic discrimination. AI systems learn from large datasets that may contain historical, social, or cultural biases. As a result, algorithms may reproduce and even amplify discrimination based on gender, age, ethnicity, or social status. Such problems are particularly visible in recruitment, credit scoring, and predictive policing systems. To reduce these risks, fairness mechanisms and regular ethical impact assessments should become standard practice.

Closely related to discrimination is the threat to privacy rights. AI systems often process enormous amounts of personal information. Facial recognition, online tracking, biometric identification, and behavioral analysis technologies may significantly interfere with individuals' private lives. Consequently, data protection measures should be implemented throughout the entire lifecycle of AI systems, and the use of personal information should be governed by strict legal requirements.

Another challenge concerns accountability and responsibility. As autonomous technologies become more advanced—from self-driving vehicles to autonomous military systems—it becomes increasingly difficult to determine who should be held responsible for mistakes or harmful outcomes. If an AI system acts independently, should responsibility lie with the developer, the operator, or the owner? Clear accountability frameworks are necessary to ensure that meaningful human oversight remains in place.

Finally, attention should be paid to the issue of unequal access to AI technologies. Wealthier countries and large corporations currently possess the majority of technological resources and expertise, while less developed regions risk becoming increasingly dependent on external technologies. This may deepen social and economic inequality on a global scale. Therefore, inclusiveness and equal access should be considered important principles of ethical AI governance.

For these reasons, international organizations such as the European Commission, UNESCO, OECD, and NIST emphasize that ethical auditing should become a mandatory practice for high-risk AI systems.

3. International Approaches to AI Regulation

Since artificial intelligence technologies operate across national borders, their regulation has become a matter of global concern. Different countries have adopted different regulatory models, but all seek to balance technological innovation with the protection of fundamental rights.

The European Union has developed one of the most comprehensive approaches through the proposed Artificial Intelligence Act (AI Act). The regulation classifies AI systems according to risk levels. Technologies that may significantly affect people's lives and rights—such as medical AI systems, facial recognition technologies, and judicial decision-support tools—are categorized as high-risk and subject to strict oversight requirements, including mandatory auditing procedures.

The United States follows a different strategy. Rather than implementing centralized federal regulation, the U.S. relies heavily on voluntary standards and industry guidelines. The National Institute of Standards and Technology (NIST) has developed the AI Risk Management Framework to promote responsible AI development. Additionally, individual states have introduced laws governing personal data protection and biometric technologies. While this flexible approach encourages innovation, it may also result in inconsistent regulatory practices across different jurisdictions.

China has adopted a more centralized model based on strong government oversight. AI technologies are viewed not only as drivers of economic development but also as tools for public security and social management. Chinese regulators have imposed restrictions on recommendation algorithms and content-distribution systems, while requiring companies to disclose certain aspects of algorithmic operation. In this model, collective interests and state security often take precedence over individual freedoms.

Russia is developing its own regulatory approach through the National Strategy for Artificial Intelligence Development until 2030. The primary focus remains on supporting research, creating technological infrastructure, and developing legal frameworks. In recent years, increasing attention has been paid to ethical principles such as transparency, fairness, and accountability. Unlike the European model based on detailed regulation or the Chinese model emphasizing state control, Russia currently pursues a more flexible and adaptive approach.

Although these models differ significantly, they share a common objective: reducing ethical and social risks while ensuring that AI technologies remain compatible with fundamental human rights and values.

4. The Russian Context and Legislative Challenges

The development of artificial intelligence in Russia is actively supported by the government through national strategies and digital transformation programs. Their primary objective is to stimulate research, develop technological infrastructure, and encourage the implementation of AI solutions across key sectors of the economy. However, the rapid pace of technological progress currently exceeds the capacity of existing legislation, which does not yet provide comprehensive regulation of AI technologies.

One of the main challenges is the absence of a dedicated law governing artificial intelligence. Existing legal provisions, including Federal Law No. 152-FZ “On Personal Data,” address only a limited range of risks associated with algorithmic systems. As a result, significant legal uncertainty remains regarding algorithm transparency, accountability, and compliance with ethical standards.

Integrating ethical principles into technological products is especially difficult. At present, Russia does not have formalized procedures for independent ethical audits or mandatory assessments of AI systems for potential discrimination and bias. This increases the likelihood of opaque and unfair algorithmic decisions. At the same time, AI applications in healthcare, finance, and public administration require a high level of public trust, which cannot be achieved without transparent oversight mechanisms.

For example, in healthcare, Russia is actively implementing telemedicine services and AI-powered clinical decision-support systems. Algorithms are already used to analyze X-ray images and assist in diagnosing pneumonia and oncological diseases. These technologies significantly improve efficiency and support medical professionals. However, when the decision-making process remains opaque, patients may question the reliability of diagnoses and become less willing to trust the healthcare system. A similar situation exists in the financial sector. AI-based systems are widely used to assess creditworthiness and determine lending decisions. If applicants receive loan rejections or reduced credit limits without a clear explanation, trust in financial institutions decreases, and companies may face accusations of unfair treatment or discrimination. In public administration, algorithms are increasingly employed to distribute social benefits and automate administrative procedures. Any error or unjustified decision that cannot be independently verified may undermine public confidence in government institutions and generate significant social concern.

Personal data protection is another major challenge. Although Federal Law No. 152-FZ regulates the processing of personal information, modern AI systems often operate beyond the scope of traditional data-processing models. The current legislation does not adequately address several critical issues.

· First, the law does not establish clear requirements for AI systems that make legally significant decisions without direct human involvement, such as loan approvals or the allocation of social benefits. Furthermore, procedures for appealing such decisions remain insufficiently developed.

· Second, the legislation was designed primarily for conventional databases, whereas contemporary AI systems process a wide range of information, including biometric data, facial images, voice recordings, geolocation data, and behavioral patterns, often in real time.

· Third, the law does not adequately regulate situations in which previously collected personal data are reused for training new AI models or transferred to third parties without fully informed user consent.

· Fourth, there are no specific legal provisions aimed at preventing algorithmic discrimination against particular social, ethnic, or demographic groups.

Finally, existing regulations focus primarily on the collection and processing of personal data rather than requiring transparency regarding the logic behind algorithmic decisions. Consequently, the current legal framework does not provide a sufficient level of oversight for AI systems that process personal information and directly affect citizens’ rights. Addressing these shortcomings will require substantial adaptation of legislation to reflect the unique characteristics of algorithmic decision-making systems.

The situation is further complicated by a shortage of specialists who are equally versed in technology, law, and ethics. The education and training of such professionals should become an important component of Russia’s digital transformation strategy.

Thus, Russia currently combines strong technological potential with an underdeveloped legal and ethical framework for AI governance. To ensure the safe integration of AI technologies, a comprehensive approach is required, including algorithm transparency, independent auditing mechanisms, and the preparation of qualified experts. Only under these conditions can risks be minimized and public trust in digital technologies be strengthened.

5. Proposals for Improving Legal Regulation

The effective development of artificial intelligence requires a comprehensive legal and ethical framework capable of maintaining a balance between technological innovation and the protection of human rights. An analysis of international experience and Russian practice makes it possible to identify several priority areas for improvement.

First, a dedicated law specifically addressing artificial intelligence should be adopted. Such legislation should establish fundamental principles governing the development and deployment of AI systems, including transparency, accountability, ethical compliance, data protection, and responsibility for the consequences of algorithmic decisions. A dedicated legal framework would provide greater certainty for developers, businesses, regulators, and citizens. In addition, legislation should establish clear requirements regarding the transparency and explainability of AI systems. Users should be able to understand the basis on which algorithmic decisions are made. Greater transparency would increase the legitimacy of AI technologies and reduce public distrust.

Second, it is necessary to introduce mandatory ethical auditing, particularly for high-risk AI systems operating in healthcare, finance, law enforcement, and public administration. Independent evaluations of algorithms for potential bias, discrimination, and compliance with legal standards would significantly reduce social risks and increase public confidence in AI applications.

The experience of the European Union demonstrates that ethical auditing can become an effective tool for ensuring responsible AI deployment. Similar mechanisms could be adapted to the Russian legal environment while taking into account national specificities.

Third, special attention should be devoted to the preparation of professionals who possess interdisciplinary expertise in technology, law, and ethics. Universities and educational institutions should develop programs focused on AI governance, ethical auditing, and digital regulation. Such specialists will play a key role in establishing standards, conducting audits, and supporting the responsible implementation of AI systems.

Another important direction involves the harmonization of Russian legislation with international standards and best practices. Compatibility with regulatory approaches adopted in the European Union, the United States, and other technologically advanced countries would facilitate international cooperation and improve the global competitiveness of Russian AI technologies.

Furthermore, certification mechanisms should be introduced for high-risk AI systems. Similar to quality and safety certifications in other industries, AI certification could serve as evidence that an algorithm has successfully passed technical, legal, and ethical evaluations.

The establishment of independent oversight bodies could also contribute to more effective governance. Such organizations could monitor compliance with ethical standards, investigate complaints, conduct audits, and provide recommendations regarding the safe use of AI technologies.

Overall, improving AI regulation should involve several interconnected measures:

• adoption of a dedicated AI law;

• implementation of mandatory ethical auditing and certification procedures;

• development of educational programs for AI governance specialists;

• enhancement of transparency and explainability requirements;

• harmonization with international regulatory standards;

• creation of independent oversight mechanisms.

Such a comprehensive approach would maximize the benefits of AI while minimizing its ethical, social, and legal risks. It would also help establish a trustworthy environment in which technological innovation can develop responsibly and sustainably.

Conclusion

Artificial intelligence is becoming an increasingly important driver of economic growth, healthcare development, educational innovation, and public administration. At the same time, the rapid expansion of AI technologies creates significant ethical, social, and legal challenges that require careful consideration and effective regulation. One of the central concerns is the “black box” nature of many AI systems, which makes it difficult to understand how decisions are generated. This lack of transparency can reduce public trust, complicate accountability, and increase the risk of unfair or discriminatory outcomes. As a result, the principles of explainability, transparency, and accountability have become essential components of responsible AI development.

The analysis conducted in this study demonstrates that ethical auditing can serve as an effective mechanism for identifying algorithmic bias, preventing discrimination, protecting personal data, and ensuring compliance with fundamental human rights. Ethical audits allow organizations to detect risks before AI systems are deployed on a large scale and help create technologies that are more trustworthy and socially responsible.

International experience shows that different countries have adopted different approaches to AI regulation. The European Union relies on comprehensive legal regulation through the AI Act, the United States emphasizes voluntary standards and risk-management frameworks, while China implements a more centralized model of state oversight. Despite these differences, all approaches share a common objective: ensuring that technological progress remains consistent with ethical principles and human values.

In Russia, the development of artificial intelligence is progressing rapidly, but the legal and ethical framework remains insufficiently developed. Existing legislation does not fully address issues such as algorithmic transparency, automated decision-making, ethical auditing, and accountability for AI-related harms. Therefore, the creation of a comprehensive regulatory framework should become a priority for future policy development.

This study concludes that the successful integration of artificial intelligence into society is only possible through a balanced approach that combines support for innovation with the protection of human rights, ethical standards, and public interests. The introduction of mandatory ethical audits, improved transparency requirements, certification mechanisms, and specialized educational programs would significantly contribute to the responsible development of AI technologies.

Ultimately, ethical auditing should not be viewed as an obstacle to innovation but rather as an essential tool for ensuring that artificial intelligence serves society in a fair, transparent, and beneficial manner. The future success of AI will depend not only on technological capabilities but also on society’s ability to establish effective ethical and legal safeguards.

References

1. Amazon scraps secret AI recruiting tool that showed bias against women. (2018). Reuters. https://www.reuters.com/article/idUSKCN1MK08G

2. Angwin, J., Larson, J., Mattu, S., & Kirchner, L. (2016). Machine bias: There's software used across the country to predict future criminals. And it's biased against blacks. ProPublica. https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing

3. Buolamwini, J., & Gebru, T. (2018). Gender shades: Intersectional accuracy disparities in commercial gender classification. Proceedings of Machine Learning Research, 81, 1–15. https://proceedings.mlr.press/v81/buolamwini18a/buolamwini18a.pdf

4. European Commission. (2019). Ethics guidelines for trustworthy AI. https://digital-strategy.ec.europa.eu/en/library/ethics-guidelines-trustworthy-ai

5. National Institute of Standards and Technology. (2023). Artificial intelligence risk management framework (AI RMF 1.0). U.S. Department of Commerce. https://www.nist.gov/itl/ai-risk-management-framework

6. OECD. (2019). OECD principles on artificial intelligence. https://oecd.ai/en/ai-principles

7. Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science, 366(6464), 447–453. https://doi.org/10.1126/science.aax2342. https://pubmed.ncbi.nlm.nih.gov/31649194/

8. European Commission. (2021). Proposal for a regulation of the European Parliament and of the Council laying down harmonized rules on artificial intelligence (Artificial Intelligence Act). EUR-Lex. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:52021PC0206

9. UNESCO. (2021). Recommendation on the ethics of artificial intelligence. https://unesdoc.unesco.org/ark:/48223/pf0000381137

10. Federal Law of the Russian Federation No. 152-FZ. (2006, July 27). On personal data (as amended July 24, 2023). https://www.consultant.ru/document/cons_doc_LAW_61801/

Competing interests

The author declares that they have no competing interests.

Use of Artificial Intelligence (AI)

The author declares that they did not use generative AI to come up with new ideas for their review.