Artificial intelligence now depends as much on physical infrastructure—such as datacenters, advanced chips, high-performance networking, and cloud platforms — as it does on models and algorithms. As Bahrain and its Gulf neighbors invest rapidly in AI infrastructure, the question of whose technology underpins it and how the AI supply chain is secured has become central to both national security and economic prosperity.

This project examines the emergence of “dark LLMs,” AI models marketed or shared on the dark web for activities such as phishing, malware development, fraud, and other forms of cybercrime. It maps this growing ecosystem, develops a typology of these tools and their intended uses, and assesses whether they are meaningfully transforming the accessibility, scale, and sophistication of cybercriminal operations.

The AI Infrastructure & Society Initiative examines the material foundations of artificial intelligence, including where computing infrastructure is built, whose land, water, and energy it consumes, and who participates in the decisions governing its expansion. Combining large-scale empirical analysis with critical infrastructure research, the initiative equips policymakers and civil society to address the environmental, social, and distributional consequences of the AI buildout.

In collaboration with the Center for Democracy & Technology and the NYU Stern Center for Business and Human Rights, this project examined how trust and safety practices in online games can inform the governance of digital platforms more broadly. The collaboration brought together experts from industry, civil society, government, and academia for a public workshop and produced a series of five essays and a white paper on pro-social design, content moderation, privacy, child safety, and community governance.