Brown wrapping paper being torn away to show two letters on a keyboard: 'AI' Photo: By Immo Wegmann on Unsplash.
Data integrity: Andrew Nind reflects on AI and the Quaker testimonies
‘Quaker testimonies can guide our response to the AI stampede.’
On 29 January, the government published a policy paper, ‘AI Opportunities Action Plan: One year on’, which features some bold – and entirely positive – claims about the benefits that AI will bring. The government claims that, in that one year, it has ‘moved decisively from ambition to delivery’ by starting ‘the biggest skills drive in a generation… upskilling 10 million workers by 2030’ and ‘supercharging the economy’. In the state sector, it is ‘embedding AI into frontline services… creating thousands of jobs… [and expanding] public sector compute [sic] 20x’ by 2030.
In the face of this one-sided rhetoric, it is instructive to consider how our Quaker testimonies can guide our response to the AI stampede. This is a big topic, so here I want to limit our focus to environmental considerations, leaving a more general analysis for a future date.
At the Just Algorithms Action Group (JAAG), we consider each Quaker testimony in turn, and how it can bring insight. Some of this is quite subtle, but let’s start with the most obvious point, sustainability. We are the guardians of the natural world. So it is alarming that, in September 2025, the Government Digital Sustainability Alliance (GDSA) predicted that, as early as 2027, global AI water demand could reach 6.6 billion cubic metres – equivalent to about half of the UK’s entire water usage. And over the course of 2026, it predicts that global AI electricity demand might exceed 1,000 TWh, more than twice as much as in 2022 – and equivalent to three times the UK’s current power consumption.
But while AI’s thirst for water and electricity is alarming, what can we do about it? Well, the close correlation between quantity of data and resource consumption suggests that we could usefully consider whether we need so much data.
We are led to believe in the virtues of ‘big data’, but what we need is an Occam’s Razor-approach to algorithms and data requirements. Rather than making algorithms more complex, or trying to plug gaps in data, it is sometimes better to ask: is all of it necessary? A range of techniques, including curve-fitting, statistical methods, and principal component analysis, can be used to constrain computer consumption of energy.
In communications with policymakers, we can all challenge the need for this data without being AI experts. We can draw attention to Frugal AI, a British university-backed initiative to develop AI with minimal resource use. And in our own lives, we can limit our reliance on AI applications like ChatGPT and its successors. We can explore the internet without getting AI-generated summaries, by adding ‘-ai’ to the end of any search query. These examples illustrate that we can influence our personal AI-related impact on the world.
And so to truth. Consider the following example, one among many. On 18 February 2026, Microsoft announced that it had achieved an important milestone in its journey to become ‘carbon negative’ by 2030. The company claimed it had succeeded in matching 100 per cent of its annual global electricity demand with renewable energy, implying that the greenhouse gas emissions from its electricity use are now ‘net zero’. But the truth testimony calls us to seek out the whole truth in ourselves and others, and to be wary of partial disclosure. Microsoft may indeed be able to show that it has signed commercial contracts for 40GW of renewable energy capacity, and that the expected generation from this capacity matches the company’s annual electricity consumption. But what about all that is not said? The environmental impact from the construction of data centres and associated infrastructure? The use of resources such as steel and rare earth minerals? How much of that renewable energy capacity would have been developed anyway? The answer is likely to be: most of it. By and large, a company like Microsoft is not increasing the world’s supply of green electricity, it is just signing up to get the electricity generated from commercially-attractive projects before someone else does. Microsoft is, however, adding to electricity demand when it builds a data centre. If demand goes up, but green generation is the same with or without the data centre, then the net effect must be an increase in electricity generated from fossil fuels.
‘We need an Occam’s Razor-approach to algorithms and data requirements.’
The next example highlights another link to our testimonies: equality. Large tech companies are effectively jumping the queue in taking credit for the development of the world’s renewable energy sources. The world’s brown energy – all the gas-fired and coal-fired power stations around the globe – is in consequence disproportionately attributed to those with a weaker voice, namely the world’s poor, who in effect take the blame for more carbon emissions as the tech companies’ measured footprint goes down.
As Quakers, we should call for fairer measurement, more rigorous regulation, and verifiable investigations into green claims. This is not just a question of fair reputation; there are practical implications. If our politicians were to recognise that having fewer and smaller data centres is better from an environmental perspective, they might be less gung-ho about AI expansion.
Finally, to peace. On 9 April 2025, the Guardian published a piece of investigative journalism on three Amazon data centres in northern Spain. They are licensed to use an estimated 755,720 cubic metres of water a year, enough to irrigate 230 hectares of corn. Water is heavy and difficult to transport. If it is being diverted to cool a data centre, it is unlikely to be transported back again to irrigate agricultural land. Instead, after serving its cooling function, it will either evaporate or be flushed away. Such examples often go unreported. Western media pays less attention to similar impacts in poorer parts of the world than Spain. One can imagine a succession of events cascading into violence. First, a new data centre siphons off water in one of the world’s driest areas. Second, food and water shortages affect civilians. Third, violence erupts. Our peace testimony calls us to speak out against such appropriation of precious resources.
There are some AI applications that are environmentally beneficial, such as the automation of heating and lighting controls. But their proponents often omit the possibility that people could still reduce their consumption without these applications. The preceding examples show there are plenty of causes for concern, notwithstanding the one-sided government narrative; hopefully they provide a taster of ways in which AI might be regulated, and how we Quakers might engage in public discourse.
Our testimonies are a powerful tool for discernment and action. This article covers just one subset of links between them and AI. Hopefully as we discern further we can explore more.
Andrew is from the Just Algorithms Action Group.
