Global AI R&D divide
AI-driven concentration of power and resources within certain entities or groups, especially those with access to or ownership of powerful AI systems, leading to inequitable distribution of benefits and increased societal inequality.
"Large companies in countries with strong digital infrastructure lead in general- purpose AI R&D, which could lead to an increase in global inequality and dependencies. For example, in 2023, the majority of notable general- purpose AI models (56%) were developed in the US. This disparity exposes many LMICs to risks of dependency and could exacerbate existing inequalities."(p. 119)
Supporting Evidence (2)
"The rising cost for developing general- purpose AI is the main reason for this ‘AI R&D divide’. Access to large and expensive quantities of computing power has become a prerequisite for developing advanced general- purpose AI. Academic institutions and most companies, especially those in LMICs, do not have the means to compete with large tech companies."(p. 119)
"A main driver of the AI R&D divide is the difference in access to compute between different actors. This includes the unequal access to powerful computing resources (graphics processing units (GPUs), data centres, cloud services, etc.) that are necessary to train and deploy large and complex AI models. In recent years, this divide has widened (721, 722)."(p. 120)
Other risks from Bengio2025 (13)
Risks from malicious use
4.0 Malicious Actors & MisuseRisks from malicious use > Harm to individuals through fake content
4.3 Fraud, scams, and targeted manipulationRisks from malicious use > Manipulation of public opinion
4.1 Disinformation, surveillance, and influence at scaleRisks from malicious use > Cyber offence
4.2 Cyberattacks, weapon development or use, and mass harmRisks from malicious use > Biological and chemical attacks
4.2 Cyberattacks, weapon development or use, and mass harmReliability issues
7.3 Lack of capability or robustness