Transnational AI regulation needed to protect human rights in the UK

The recent inquiry into the regulation of artificial intelligence (AI) in the UK has underscored the urgent need for transnational regulatory frameworks to address the complex safety issues posed by advanced AI technologies. During the final evidence session of the Joint Committee on Human Rights held on February 25, MPs and Lords engaged with AI minister Kanishka Narayan and senior executives from tech giants Meta and Microsoft, probing the adequacy of current safeguards designed to protect fundamental human rights as AI systems become increasingly integrated into everyday life.

The discussions revealed a growing concern among lawmakers regarding various aspects of AI governance, including misinformation, accountability, child safety, and the existential risks associated with AI technologies. The committee's inquiry comes on the heels of warnings that the UK’s existing regulatory framework is struggling to keep pace with the evolving harms posed by AI, with several regulators citing a lack of resources as a significant barrier to effective oversight rather than a deficiency in statutory powers.

Fragmentation of Regulation

Ginny Badanes, the general manager of tech for society at Microsoft, and Rob Sherman, the deputy chief privacy officer of policy at Meta, both advocated for greater harmonization of regulatory standards on a global scale. Badanes articulated concerns about the fragmentation of regulatory approaches, stating, "I worry at times when we have this variety of approaches that we’re not actually addressing the broader safety or human rights risks that are at the centre of what everyone is trying collectively to solve." She emphasized that the inherently transnational nature of advanced AI necessitates a unified regulatory framework, as these systems are developed and deployed across multiple jurisdictions and supply chains.

Sherman echoed this sentiment, highlighting that Meta operates globally and does not create separate AI models for different countries, despite the variations in governance. Both executives expressed support for the UK’s AI Opportunities Action Plan, which they believe strikes a reasonable balance between fostering innovation and protecting human rights. Badanes noted that public trust is crucial for the adoption of AI technologies, asserting that strong yet proportionate regulations can help secure that trust.

Accountability and Trust in AI

The inquiry also delved into the critical issue of accountability in the context of AI-generated outputs. When questioned about the responsibility for harm resulting from incorrect or manipulative AI outputs, both executives refrained from proposing a specific legal framework. Badanes suggested that accountability should be linked to where there is meaningful control, indicating that responsibility could vary based on whether harm arises from the AI model itself, its deployment, or misuse by malicious actors. Sherman acknowledged that courts would likely need to consider multiple stakeholders in such cases.

The committee raised concerns about the reliability of large language models, which can produce false information, commonly referred to as "hallucinations." While these models are designed to provide accurate information, both executives admitted that achieving consistent truthfulness is a complex challenge due to the inherent flaws in their design. They also emphasized the importance of AI literacy and managing public expectations regarding the capabilities of chatbots and AI systems.

Badanes pointed out that the risk associated with AI varies significantly across different applications, advocating for a risk-based approach to governance rather than a one-size-fits-all regulatory model. She argued for the necessity of building age-appropriate designs and safety measures into AI systems from the outset, rather than merely restricting access.

The Path Forward for AI Regulation

The inquiry concluded with a broader discussion about the UK's strategy for enhancing its AI sovereignty. Minister Narayan outlined a three-step model aimed at securing critical inputs, diversifying suppliers, and building domestic capabilities in AI. He acknowledged the limitations of the UK’s control over the entire AI ecosystem but highlighted strengths in areas such as chip design and AI for scientific applications.

Badanes drew parallels between AI regulation and nuclear regulation, suggesting that society has successfully navigated similarly complex challenges in the past. However, concerns were raised about the departure of AI researchers from major companies over safety disagreements, prompting questions about the sufficiency of voluntary corporate safeguards. Sherman noted that while firms have internal reporting mechanisms and encourage dissent, he stopped short of endorsing binding global treaties.

As the inquiry wrapped up, industry leaders urged policymakers to prioritize the establishment of interoperable, risk-based global standards for advanced AI systems and to invest in tools that enhance content provenance to combat misinformation. The dialogue emphasized the need for collaborative efforts between the tech industry and governments to ensure that AI technologies are developed and deployed in a manner that safeguards human rights and democratic values.

In summary, the inquiry into AI regulation in the UK has highlighted critical issues surrounding the need for transnational regulatory frameworks, accountability, and public trust in AI technologies. As AI continues to evolve and permeate various aspects of society, the call for cohesive and effective governance becomes increasingly vital to protect fundamental rights and freedoms.