Last year, the United Arab Emirates (UAE) made global headlines with the release of Falcon, its open-source large language model (LLM). Remarkably, by several key metrics, Falcon managed to outperform or measure up against the LLMs of tech giants like Meta (Facebook) and Alphabet (Google).
Since then, the UAE has positioned itself as a frontrunner in the global artificial-intelligence race by consistently releasing updates to its powerful model. These efforts have not gone unnoticed: In April, Microsoft acquired a $1.5 billion minority stake in G42, the UAE’s flagship AI company, underscoring the country’s growing influence.
Analysts often attribute the UAE’s emergence as an AI powerhouse to several factors, including robust state support, abundant capital, and low-cost electricity, all of which are necessary for training LLMs. But another important – and often overlooked – factor is the country’s authoritarian governance model, which enables the government to leverage state power to drive technological innovation.
The UAE is not alone. Authoritarian countries like China have a built-in competitive advantage when it comes to AI development, largely owing to their reliance on domestic surveillance, which fuels demand. Facial-recognition technologies, for example, are used by these regimes not just to enhance public safety but also as powerful tools for monitoring their populations and suppressing dissent.
By contrast, facial recognition has become a source of enormous controversy in the West. The European Union’s AI Act, which entered into force on August 1, has effectively banned its use in public spaces, with only a few narrowly defined exceptions.
Top-down national strategies
This provides AI firms in China and the UAE with a massive advantage over their Western counterparts. Research by David Yang and co-authors shows that Chinese AI firms with government contracts tend to be more innovative and commercially successful, owing to procurement practices that provide them with access to vast troves of public and private data for training and refining their models. Similarly, UAE firms have been allowed to train their models on anonymized health-care data from hospitals and state-backed industries.
AI firms seeking access to such data in Western countries would face numerous legal hurdles. While European and American companies grapple with strict compliance requirements and a surge in copyright-infringement lawsuits, firms in China and the UAE operate in a far more lenient regulatory environment.
You have 55.09% of this article left to read. The rest is for subscribers only.