In December 2025 I hit two personally notable milestones: the discovery of the Effective Altruism community and my 10 year anniversary of volunteering with the foodbank network in my local city. Across my ten years of foodbank volunteering, I've seen a lot of change: most significantly food support demand has increased by a factor of five, meaning we've grown from 5 loosely aligned independent foodbanks to a coordinated network of 16 hubs supported by a central warehouse and logistics network.
Today we also do a lot more than hand out food parcels: we supply furniture for council houses, run a school uniform service for those who can't afford it, offer landlord dispute legal dispute...and administer the 'phone and internet bank' service for our local council. It's the latter that has me wondering if (or more likely when) AI banks are going to need to enter the portfolio. I also wonder which organisations will drive them into being.
For those wondering what an "AI bank" means, the easiest way to understand it is through the evolution of internet banks. As the internet became wide spread and life moved online, not having access to the internet gradually became significantly barrier to accessing education, work and public services. People began to discuss whether internet access needed to something become akin to a human right.
We haven't gone that far in the UK, but the government has made a number of interventions, beginning with ensuring internet was provided in public libraries and more recently through introducing new legislation. In 2020 the Broadband Universal Service Obligation (USO) gave people the right to request a basic level of broadband capability and paved the way for the introduction of social tariffs - subsidised internet rates for people financially assessed as struggling to otherwise afford it.
As AI becomes more mainstream and becomes 'essentially integrated' within global tech and eco-systems, similar questions around equitable access to AI are probably going to arise. As Sir Francis Bacon wrote in 1597 "knowledge itself is power" while in 1748 Benjamin Franklin advised that "time is money". AI tools access combined with AI skills literacy offers the potential to create personal leverage: the ability to access and effectively navigate a much wider range of knowledge faster and to save time by utilising creative AI tools and automated skills (take a look at Claude Cowork if you haven't). But like all transformative tools if offers the potential for those who aren't empowered with them to get critically left behind and structurally disadvantaged.
AI usage is growing: a government survey found that in early 2026 1 in 3 adults had used generative AI within the last month, while the Guardian newspaper reported that 92% of a sample of 1000 students had used some form of AI during their studies. The more provocative question was why the remaining 8% hadn't.
Volunteering at the foodbank I know that one reason some of the 8% may not have is money. As the OpenAI v Musk trial is highlighting AI systems incur non-trivial data and energy costs. This is part of what is driving the rise of pay-to-use subscription tiers. Yes, many AI tools still have free tiers, but they're heavily limited and often intentionally designed as 'lead hooks' showing you just enough of a taste of the value you can only truly access by converting to a paying customer. Not everyone can afford a £20/month AI subscription though, and imagine growing up as a future child in a world where AI is theoretically everywhere but practically inaccessible to you. Enter: AI banks.
The AI Banks will need to be about than just providing free or affordably subsided access to higher tier capabilities. As was found with the internet - providing access isn't enough, you also need to provide the training on how to use it, the safeguards around it, the infrastructure to support it and organisations to manage it. In the case of AI that may mean things like like prompt engineering skills training and high performance computing access. It may mean AI companies providing access to their capability (voluntarily or enforced) and governments providing capability hubs. What the concept means in a world with AGI - well I'd love your thoughts on that.
For all this isn't an AI safety topic, I do personally think ensuring equitable access to AI and AGI is an important aspect of doing AI responsibility and I'm keen to contribute to activity and discussions on this kind of topic. Using our foodbank network leverage to raise this with local government working groups seems like one place to start, but I'd welcome other ideas and thoughts on bringing the 'effective altruism' cost effectiveness validation concepts to this idea.