AI Regulation Challenges in an Evolving Tech Landscape

Fast AI Development Creates A Regulatory Gap

The current buzz surrounding Generative AI, exemplified by innovations like ChatGPT, is, in essence, an outcome of years of technological and scientific advancements.

It is crucial to acknowledge that these breakthroughs are not sudden; rather, they are the culmination of ongoing progress and development in the field.

The future of LLMs is open source decentralization. The harder the centralised push for regulation and control, the faster the acceleration. AI projects like llama2 by @ylecun lead the way. https://t.co/d2iF0fgET6

— Teodor Mitew (@tedmitew) November 1, 2023

AI expert Maurizio Marcon said : “Today, a long-standing issue that is becoming increasingly urgent is the regulation of technological progress,”

He added: “Governments and legislative bodies have operational times that are incompatible with the speed at which innovations are now made available on the market. This can lead to regulatory gaps that can be exploited, creating problems for society and the economy.”

Challenges Of Regulating An Evolving Technology

Regulating AI is a complex and challenging task for several reasons. AI is a rapidly evolving technology, with new advancements constantly being made. This makes it difficult to develop and implement regulations that are both effective and adaptable.

Velocity

Preventing the corporate AI race from descending into recklessness necessitates the creation and evolution of rules, coupled with the rigorous enforcement of legal guardrails. Nevertheless, coping with the rapid pace of AI-driven transformation can surpass the current expertise and authority of the federal government. The regulatory statutes and structures currently at the government’s disposal were based on assumptions from the industrial era, and they have already been surpassed by the initial decades of the digital era. The existing regulatory framework lacks the necessary agility to effectively cope with the swift pace of AI development.

Former Google Executive Chairman and current AI evangelist Eric Schmidt has warned : “There’s no one in government who can get it [AI oversight] right.”

While Schmidt recognizes the need for behavioral expectations, he explains : “I would much rather have the current companies define reasonable boundaries.”

This self-regulatory stance mirrors the “leave us alone” approach championed by digital platform companies over the past two decades. The outcomes of such a strategy are evident in well-documented contemporary online harms. These include the unparalleled invasion of personal privacy, market consolidation, manipulation of users, and the widespread dissemination of hate, false information, and misinformation.

In the realm of AI, a more robust solution is imperative, as corporate self-regulation may prove insufficient. This is evident when the pursuit of profits is anticipated to outpace the establishment of meaningful safeguards.

What To Regulate?

Given the multifaceted nature of AI, a blanket “one-size-fits-all” regulatory approach runs the risk of over-regulating certain instances while under-regulating others. For example, the impact and implications of AI in a video game differ significantly and warrant distinct treatment compared to AI applications that could pose threats to critical infrastructure or human safety.

Consequently, effective AI regulation must adopt a risk-based and targeted approach that acknowledges and addresses the specific risks associated with diverse AI applications.

Who Has To Regulate?

In the ongoing digital age in the United States, rule-making has largely been the purview of innovators due to the government’s historical failure to establish comprehensive regulations. This outcome is unsurprising, given that rules developed by the industry naturally tend to favor their creators. Acknowledging the imperative for AI policies, the pivotal question arises: who should be responsible for formulating these policies?

During his testimony in May this year, Sam Altman, the CEO of OpenAI endorsed the notion of a dedicated federal agency overseeing AI. Previously, Brad Smith of Microsoft and Mark Zuckerberg of Meta have also supported the idea of a federal digital regulator.

Just two days after Altman’s hearing, Senators Michael Bennet (D-CO) and Peter Welch (D-VT) introduced a bill. This proposed the establishment of a Digital Platform Commission (DPC). This bill not only proposed the creation of a new agency with the authority to address challenges posed by digital technology, including AI. But also advocated for an agile, risk-based approach to regulatory development. Reportedly , Senators Lindsey Graham (R-SC) and Elizabeth Warren (D-MA) are working on their proposal for a digital agency.

The U.S. Congress must mirror the innovators’ expansiveness and creativity from the digital revolution while considering the establishment and operations of a new agency for the evolving digital landscape, encompassing AI.

The Role of International Cooperation

The challenges of regulating AI are not confined to any one jurisdiction. As AI becomes more widely used, it will be increasingly important for countries to work together to develop effective regulatory frameworks.

The European Union has been a leader in the development of AI regulations, with its proposed Artificial Intelligence Act (AIA). The AIA aims to establish a comprehensive framework for regulating AI in the EU. It covers a wide range of issues, including fairness, transparency, and accountability.

The US is actively addressing AI regulations. The Department of Commerce issued a 2022 report on AI’s potential harms and benefits. The report advises adopting a risk-based regulatory approach, emphasizing heightened oversight for AI systems posing substantial harm.

What Are Governments Doing?

Governments worldwide are actively proposing legislation, guidance, and regulations to delineate the appropriate use of AI. Notably, the European Union’s AI Act is the inaugural AI legislation globally. Countries such as Canada , the United States , and the United Kingdom are forging their legal frameworks. Simultaneously, 46 nations have voluntarily embraced the Organization for Economic Co-operation and Development’s AI Principles—non-binding yet profoundly influential guidelines.

The EU’s AI Act is a strong step in the right direction for human creators.

It’s not the part about complying with copyright law that’s most impactful – companies will continue to argue over how copyright law should be interpreted.

The key part is that foundation models will… pic.twitter.com/FV5dMbkqoU

— Ed Newton-Rex (@ednewtonrex) December 10, 2023

Amidst these efforts, States are delicately navigating the balance between ensuring the safe development of AI and fostering innovation. Even in regions where some regulation, or at least the potential for regulation, exists, a form of self-regulation can manifest regionally. This occurs when corporations engage in “regulator shopping,” strategically choosing jurisdictions that offer more favorable treatment than others. If governments observe such maneuvers, a potential race to the bottom may unfold. This could prompt them to relax regulations in a bid to attract investment.

This dynamic risks companies exporting their AI governance values to areas with limited alternatives. While these exports may conform to global norms, they jeopardize the legitimacy of locally crafted AI regulations.

A Possible Solution?

The rapid pace of AI development has created a significant gap between the capabilities of AI systems and the ability of existing laws and regulations to govern them effectively. This gap poses significant risks. Among these, is the potential for AI systems to be used to discriminate against individuals, invade privacy, and manipulate behavior.

To address these risks, it is essential to adopt a multi-pronged approach to regulation that encompasses both traditional legal frameworks and new approaches tailored to the specific characteristics of AI. This approach should include:

Enact clear and comprehensive legislation that establishes general principles for the development, deployment, and use of AI systems. This legislation should address issues such as algorithmic bias, transparency, accountability, and human oversight.

Develop sectoral regulations that address specific AI applications in areas such as healthcare, finance, and autonomous vehicles. These regulations should be tailored to the unique risks posed by each application.

There's a big catch in the EU's landmark new AI law https://t.co/NdSyNi6gU4

— Axios (@axios) December 12, 2023

Establish independent regulatory bodies with expertise in AI to oversee the development and deployment of AI systems. These bodies should have the authority to conduct audits, issue compliance orders, and impose penalties for violations of AI regulations.

Promote international cooperation on AI regulation to ensure that AI systems development and deployment occur in a responsible and safe manner across borders. This cooperation could involve the adoption of common standards, the exchange of information, and the coordination of enforcement efforts.

Support research on the ethical and social implications of AI to inform regulatory decisions and public discourse. This research should focus on issues such as fairness, accountability, and the potential for AI to exacerbate existing social and economic inequalities.

AI Outpaces Legislation: Regulations Challenges in a Rapidly Evolving Tech Landscape