Unintended consequences of AI craze
At a first glance the potential environmental impact of AI development and implementation has nothing to do with online child sexual exploitation (OCSE). However, as discussed in the Part 2: Unique patterns of OCSE in Asia Pacific: AI will not be enough of the series, OCSE is a complex and a nuanced problem with numerous contributing factors, including the environmental ones.
Some projections suggest that computers could consume anywhere between 8% and 21% of global energy by 2030, which would constitute a significant increase from 1% in 2018. In addition to contributing further to the existing energy crisis, there is also a concern about the operational carbon emissions from computation. According to the International Energy Agency (IEA), “combined electricity use by Amazon, Microsoft, Google, and Meta more than doubled between 2017 and 2021”. While IEA notes that since 2010 data emissions have increased only slightly, despite a surge in digital service demand, emissions still need to drop by half by 2030 in order to get on track with the Net Zero Scenario.
This suggests that the improved energy efficiency and ICT companies buying renewable energy may not be sufficient in the face of growing demand on computational resources that are vital for training and implementation of AI solutions. Google estimates that machine learning accounted for 10-15% of its total energy use in 2019-2021, growing at a rate of about 20-25% per year. To meet this demand companies will likely continue to rely, at least partially, on fossil fuel generating high rates of CO2 and methane emission. Emissions have long been claimed to be one of the major contributing factors of climate change which in turn triggers climate-related hazards and natural disasters.
Adverse climate events produce long-term and far-reaching consequences for already marginalised and vulnerable populations. Communities that are affected by natural disasters are more vulnerable to loss of income, food and shelter and have been found to commit violence against children more frequently. With humanitarian crisis recognised as one of the known risks associated with being a victim of child sexual exploitation, any factor, in this case heavy investment in development of AI-powered tools, that contributes, either directly or indirectly, to the increase in human exposure to extreme weather events, increased flooding and droughts ought to be examined with utmost care.
With risks and benefits of new technology not being distributed equally across the globe, the countries and regions that tend to lead in AI development are not the countries that would likely benefit the least from its usage or suffer the unintended consequences it has on the environment. According to the 2022 Red Cross world disaster report the Asia Pacific region is disproportionately affected by natural disasters both in terms of the number of deaths and the total number of people affected. Of the top ten countries worst affected by disasters in 2020–2021, five are located in the Asia Pacific region - India, Indonesia, China, Pakistan, and the Philippines. What’s more, the number of climate- and weather-related disasters continues to grow, putting further populations at risk despite the notable improvements in disaster response.
As stated in Part 1 and Part 2 of this series, AI-powered tools can be an undeniable asset in the fight against OCSE. However, it’s also evident that most AI tools are intended for those who already have the most privilege in society. Implementation of AI ought to rely on a thorough risks & benefits analysis to ensure that negative consequences do not outweigh the positive impact. If AI is to be used for detection of OCSE it has to take precedence over other, perhaps more commercially-driven, initiatives that require equal or higher energy investment.
Proliferation of informal settlements
In addition to being one of the global epicentres for natural disasters, Asia Pacific also has the highest prevalence of informal settlements, often called slums, that signal economic instability. According to UN Habitat, in 2020, over 1 billion urban residents lived in such conditions, with 63% concentrated in Central and Southern Asia, Eastern and South-Eastern Asia. This translates to over 650 million people in the region. UNICEF estimates that 350 to 500 million children worldwide live in slums with limited access to vital services like healthcare, education, and sanitation.
Slums are often challenging to police due to their chaotic nature. Research indicates that criminal activities, including rape, robbery, and theft, are prevalent in slum districts due to their enabling environments and these crimes disproportionately affect women and children. Slum populations also tend to be among the most affected by the adverse weather conditions which exacerbates economic and environmental pressures further. These circumstances combined with a higher prevalence of and a higher tolerance for crimes against women and children due to the underlying social norms create an environment where commercial child sexual exploitation, including streaming on-demand content and selling of self-produced content, can be seen as a viable way to earn a living.
A nuanced evidence-based approach to OCSE detection, and more importantly, its prevention would theoretically take into account the regional patterns of offending, including the impact of environmental factors on prevalence of illicit content. Unfortunately, when mechanisms for scaled detection are being deployed they tend to target broad patterns of abuse and focus less on prevention and mitigation practices, often due to resource constraints.
Chapter conclusion
The burgeoning development and implementation of AI bring to light a myriad of unintended consequences. Despite the promises of the eventual levelling and then decrease in the AI energy demand as well as strides towards energy efficiency and renewable energy adoption by ICT companies, the demand for computational resources and energy consumption remains high. This, in turn, exacerbates the existing energy crisis, contributes to carbon emissions and threatens to intensify environmental degradation resulting in the increasing volume and severity of adverse climate events.
The disproportionate burden of these consequences falls on regions like Asia Pacific and on numerous at-risk communities that exist here. Areas, where economic instability and environmental pressures are high, create fertile ground for criminal activities, including OCSE.
Therefore, as AI inevitably continues to be integrated into society, it is imperative to conduct comprehensive risk-benefit analyses to mitigate unintended consequences. Ensuring equitable distribution of resources and prioritising initiatives that address pressing societal issues, such as OCSE, can pave the way for a more ethical and inclusive deployment of technology. Additionally, a nuanced approach that considers regional patterns of offending, including the influence of environmental factors, is essential for effective detection and prevention strategies.
See other chapters in the "Combatting OCSE in Asia Pacific" series: