## Building Blocks for AI Governance

## Source details

**Canonical URL:** [Building Blocks for AI Governance](https://www.imf.org/-/media/files/publications/fandd/article/2023/december/pov-bremmer-suleyman.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/fandd/article/2023/december/pov-bremmer-suleyman.pdf.md)
- [Structured JSON version](/-/media/files/publications/fandd/article/2023/december/pov-bremmer-suleyman.pdf.json)

---

### The AI power paradox and technological features
- AI is described as hyper-evolutionary and unpredictable; its rate of improvement "will far surpass the already powerful Moore’s Law."
- The amount of computation used to train the most powerful AI models "has increased by a factor of 10 every year for the past 10 years."
- Moore’s Law has "successfully predicted the doubling of computing power every two years"; AI improvement is faster than that, and processing that "once took weeks now happens in seconds."
- AI may become "the first technology with the means to improve on itself."
- AI proliferates easily: algorithms are easier and cheaper to copy and share (or steal) than physical assets, and as computing gets cheaper "such models will soon run on smartphones."
- Because marginal cost and marginal cost of delivery are zero, "once released, AI models can and will be everywhere."
- Single malign or "breakout" models can wreak havoc despite many models being trained responsibly.

### Geopolitical and economic incentives
- AI is a strategic objective for governments and companies; "AI supremacy is a strategic objective of every government and company with the resources to compete."
- The dynamic is framed as zero-sum between major powers: "Both the US and China see AI supremacy as a strategic objective that must be achieved—and denied to the other."
- Private sector control: "every aspect of AI is currently controlled by the private sector." Including private companies in governance is necessary; "governments should make companies parties to international agreements."
- Proliferation and decentralization: while a handful of large tech firms currently control AI, "the gradual proliferation of AI will bring more and more small players into the space, making governance more complicated."
- Constraining AI is harder than past dual-use technologies because AI’s low cost enables limitless deployment for civil or military use.

### Governance challenges and design imperatives
- Traditional, static policymaking structures are inadequate because AI’s unpredictability makes rules passed today potentially ineffective "in a few months, let alone a few years."
- A single governance regime is insufficient due to AI’s general-purpose nature; governance must be "targeted, risk-based, and modular rather than one-size-fits-all."
- Compliance must be robust: any regime must be "impermeable" because "just one defection from the regime could allow a dangerous model to escape."
- Governance must reconcile two aims simultaneously: stimulate innovation to solve intractable challenges and "avoid dangerous proliferation," while pursuing geopolitical advantage without initiating a new arms race.

### Five founding governance principles (as stated)
- Precautionary: "The risk-reward profile of AI is asymmetric; although there are vast benefits to AI’s potential, policymakers must guard against its potentially catastrophic downsides. The already widely used precautionary principle needs to be adapted to AI and enshrined in any governance regime."
- Agile: "AI governance must be as agile, adaptive, and self-correcting as AI is fast-moving, hyper-evolutionary, and self-improving."
- Inclusive: "The best industry regulation, especially when it comes to technology, has always worked collaboratively with the commercial sector ... the only way for regulators to properly oversee AI is to collaborate with private technology companies." To reflect AI’s borderless nature, "governments should make companies parties to international agreements."
- Impermeable: "Any compliance mechanisms should be watertight, with easy entry to compel participation and costly exit to deter noncompliance."
- Targeted: "Determining which tools are appropriate to target which risks will require developing a live, working taxonomy of discrete potential AI impacts."

### Democracy, participation, and AI
- AI presents risks to democratic institutions through job displacement, economic destabilization, concentration of power, and potential replacement of human decision-making with "algocracy."
- Counterpoint: societies "have the political ability (at least for now) and the responsibility to address the harm AI could inflict" and the opportunity to "harness AI to enhance our democracy."
- AI can enable more inclusive, participatory, and deliberative democracy, including at the global scale.
- Participatory experiments have been used for decades: "For 40 years many governments have engaged in experiments aiming to include ordinary citizens in policymaking and lawmaking" (e.g., citizens’ assemblies and juries on climate and other issues).
- Democratic voice should inform distributional impacts of AI and be heard at international institutions.

*Ian Bremmer and Mustafa Suleyman; "Building Blocks for AI Governance," F&D DECEMBER 2023*

---


_Source: https://www.imf.org/-/media/files/publications/fandd/article/2023/december/pov-bremmer-suleyman.pdf_