Despite AI’s wild-fire spread across sectors, AI governance and regulation is still in its nascent stages. This creates massive gaps in AI oversight and society’s ability to determine two key things:
- who benefits from this reality-altering technology, and
- who is responsible for its outputs and outcomes.
What this means and how we can ramp-up this essential AI oversight—despite considerable barriers—is the focus of Nir Kshretri’s recent Computer article, “Economics of AI Governance.”
AI Governance: Where It Stands (or Tries To)
AI governance can occur anywhere from the firm to the international level. Because AI technologies and developers—including multinational corporations—transcend national borders, international governance of AI is crucial.
Among the international bodies undertaking AI governance initiatives are the
- United Nations, which has established a 39-member advisory body—including tech executives, academic experts, and government officials from a range of countries—to address global AI governance.
- Group of Seven (G7) announced the Hiroshima AI Process in October 2023, establishing international guiding principles and a voluntary code of conduct for AI developers.
As Kshreti details, various national and regional governments are also developing or implementing governance frameworks for AI, including China, the European Union, Japan, the United Kingdom, and the United States.
While the EU and China have been aggressive in creating regulatory frameworks, most governing bodies have instead relied on “nudges and soft norms.” To fill this regulatory vacuum, Kshreti identifies three categories of prescriptive normative frameworks:
- Voluntary guidelines and codes of conduct. These industry-led efforts aim to encourage responsible AI development.
- Standards. Sometimes referred to as “soft laws,” AI standards are in development in various nations and institutions, yet none are yet fully developed.
- Certification programs. These programs, which can be a valuable element in AI governance, will demonstrate how a company’s AI processes align with standards … once those standards are fully established.
Barriers to Oversight
Kshreti highlights various barriers that hinder AI regulation, including the following:
-
- Countries possess regulatory power, but they grapple with difficult tradeoffs between ensuring AI’s safety for their citizens and facilitating an environment that encourages AI innovation—and the global advantages it offers.
- Political leaders charged with regulating AI often lack a sufficient, let alone a nuanced, understanding of AI, its risks, and its benefits.
- Professional, industry, and trade associations are tasked with monitoring member conformity with normative and coercive expectations, yet we lack normative institutions of this type that focus exclusively on AI.
In the latter case, many technology institutions heartily endorse the “ethical AI goal,” but have yet to achieve consensus on how to practically ensure or operationalize this goal.
Digging Deeper
AI requires a nuanced policy agenda that not only prevents harmful proliferation but also allows for innovation and geopolitical advantage—ideally without inadvertently triggering a new global arms race. Navigating this rocky terrain is rendered even more treacherous by major tech firms, who continue to dominate the AI space and complicate governance efforts.
To explore these and other critical AI governance issues in depth—as well as access Kshreti’s bibliography for further insights—check out “Economics of AI Governance” in Computer magazine’s April 2024 issue.
To dig even deeper, join other AI experts, researchers, government officials, and enthusiasts at the international IEEE Conference on Artificial Intelligence (IEEE CAI) 5–7 May 2025 in Santa Clara, California.
In addition to showcasing the latest AI research and breakthroughs, IEEE CAI emphasizes applications and key subject areas, from sustainability and human-centered AI to issues and industry-specific applications in healthcare, transportation, and engineering and manufacturing.
