Regulation of AI – Need & Challenges – Explained Pointwise

Calls for regulation of Artificial Intelligence has emerged stronger than ever with the recent rise in the cases of misuse of AI like proliferation of deepfakes. With the rapid advancement of AI and its potential impact on society, there is a growing consensus among experts that regulation is necessary to ensure responsible and ethical use of AI technology.

What is the need for regulation of AI?

1. Bias and discrimination: AI systems can inherit biases from the data they are trained on, leading to discriminatory outcomes. For e.g. Facial recognition algorithms have been shown to have higher error rates for women and people with darker skin tones.
2. Lack of transparency: Many AI algorithms operate as black boxes, making it difficult to understand how they reach their decisions. For e.g. Medical AI system recommending a specific medical treatment but cannot explain its reasoning.
3. Privacy and data protection: AI systems rely on vast amounts of personal data, raising concerns about privacy and data protection. For e.g. Lawsuits against Silicon Valley giants for data and privacy breach in their AI systems.
4. Security risks: AI systems can be vulnerable to cybersecurity threats and attacks. For e.g. Adversarial attacks can manipulate AI models posing risks in critical domains such as autonomous vehicles or healthcare.
5. Ethical considerations: AI raises ethical questions related to the impact on jobs, social inequality, and concentration of power. For e.g. automated decision-making in hiring processes have shown to perpetuate existing biases and result in unfair outcomes.
6. Artificial General Intelligence: AGI can self-learn and go beyond human intelligence, raising concerns of predictability and security.
7. Autonomous Weapons Development: These machines have the potential to make life-and-death decisions without direct human intervention, leading to ethical dilemma regarding the value of human life.
8. Mass State Surveillance: AI, equipped to conduct facial recognition and analyze extensive data, will empower governments to maintain round-the-clock profiles of citizens. This will make dissenting against governments difficult.
9. Challenges associated with Deepfakes generated using AI: There are concerns about women safety (morphed pornographic material), liar’s dividend (an undesirable truth is dismissed as fake news) and fuelling radicalisation and violence (Fake videos showing armed forces committing ‘crimes in conflict areas’).

What are the challenges in regulation of AI?

1. Defining AI: Defining Artificial Intelligence is challenging because AI encompasses a broad and continuously evolving range of technologies, from simple rule-based systems to advanced generative and autonomous models. A narrow definition may quickly become outdated, while an overly broad one could unintentionally regulate ordinary software.
2. Rapid technological advancement: AI technologies are evolving at an unprecedented pace, with capabilities improving faster than traditional legislative and regulatory processes. Generative AI, autonomous systems and multimodal models are creating new risks and applications before regulators can fully understand their implications. Consequently, laws may become outdated soon after they are enacted, creating regulatory gaps and uncertainty.
3. Balancing innovation and regulation: AI regulation faces the challenge of protecting society from potential harms without stifling technological innovation. Overly stringent regulations can increase compliance costs, discourage start-ups and research, and reduce investments in emerging technologies. Conversely, inadequate regulation can allow privacy violations, algorithmic discrimination, misinformation and unsafe AI applications.
4. Increased costs and competition: AI regulation may impose higher compliance costs on businesses, particularly start-ups and small enterprises. Companies may need to invest in data governance, cybersecurity, audits, documentation, testing and human oversight to meet regulatory requirements. This could increase the cost of developing and deploying AI systems.
5. Accountability and liability: AI systems can make or influence decisions that significantly affect individuals, yet determining who should be held responsible when an AI system causes harm is often difficult. Developers, technology companies, deployers and end-users may all contribute to the outcome.
6. International cooperation: AI transcends national borders, making unilateral regulation insufficient. Developing consensus among different countries with varying interests and priorities is a complex task.

What is the status of regulation of AI in India and across the globe?

What should be the way forward?

1. Adopt a risk-based framework: AI applications have different levels of potential harm, so regulation should be proportionate to the risks involved rather than applying the same rules to every AI system. A risk-based framework ensures that regulation is neither too weak to protect society nor so excessive that it suppresses innovation. This will make regulation proportionate to the potential harm posed by an AI system.
2. Universal adoption of the Bletchley Declaration: The Bletchley Declaration, adopted at the first AI Safety Summit in the UK in 2023, recognised the need for international cooperation to address risks arising from frontier AI. Its wider adoption can provide a common foundation for responsible AI governance.
3. Establish comprehensive and flexible regulatory framework: The governments should develop clear guidelines and laws that address various aspects of AI, including data privacy, algorithmic transparency, accountability, and potential biases.
4. Foster international cooperation: Given the global nature of AI and its potential impact, collaboration among countries is essential. International standards and agreements should be developed to promote ethical practices and ensure consistency in regulation across borders. In this respect, the G7 Hiroshima AI Process (HAP) could facilitate discussions.
5. Encourage industry self-regulation: Given the rapid evolution of AI, government regulation alone may struggle to keep pace with emerging technologies. Industry-led codes of conduct, technical standards and voluntary safety commitments can complement formal regulation.
6. Invest in AI research and education: Governments, academic institutions, and industry stakeholders should allocate resources to R&D, and education in the field of AI. This will help create a well-informed workforce capable of addressing regulatory challenges and ensuring the safe and responsible deployment of AI technologies.

Conclusion: The objective should not be to “regulate AI out of existence”, but to “regulate risks while enabling innovation.” This will ensure that AI becomes a tool for inclusive development rather than a source of new inequalities and vulnerabilities.