Justifying AI is hard - especially for charities
| 5 min read
I'm still trying to work out how I feel about this, so come on the journey with me...
I work in the UK charity sector, and something I've been playing over in my mind the last couple of weeks is the OpenAI / Anthropic (ChatGPT/Claude) debate. Within a lot of organisations in the sector, the idea that OpenAI would think to sign a contract with the US Military was abhorrent. Meanwhile Anthropic were standing up to the establishment, and not budging on their hard lines. Dozens of my peers flocked from one to the other almost overnight.
And this wasn't just in the charity sector, and it wasn't just in the UK. Worldwide, subscriptions to Claude doubled since their "disagreement" with the Pentagon, and free users increased by 60% (and those numbers are a couple of weeks old as I write this).
And lest I sound like I'm preaching from an ivory tower, I was one of those same people who looked at the news and thought "well ok then, where's my credit card? Good thing I pay for ChatGPT monthly".
There's no such thing as a "good" mainstream AI company
But here's the thing. While it might be true that Anthropic are no longer welcome to tender for US government contacts, Claude is the model that sits behind Palantir, a business that is well known for its Gotham platform - a tool that uses AI to aggregate information from disparate data sources to provide analysts with detailed information about an individual. The technology is being used in a number of ways, including by ICE to identify areas for their raids.
A quick sidebar to note that tech companies aren't hiding their intentions when they do this. For those who didn't grow up as nerds:
-
A Palantir is an artifact from Tolkien's The Lord Of The Rings mythology, that allows users to communicate over vast distances. It was most famously used by antagonist Sauron as a surveillance tool, a means of corrupting users to his ends, and to read the minds of individuals to gain information.
-
In the 2012 film The Dark Knight, protagonist Batman uses a high tech system developed by his company to hack into an entire city's worth of mobile phones, use them as sonar devices, and build a 3D model of the entire city including it's people and what they were doing. The city in question was called Gotham.
Back in the real world, if you think that you can switch AI models for something like a military strategy workflow overnight, I've got some bad news for you. Claude has been embedded in US military operations since July, and has been in official use by US government departments since 2024. It's going to take time to undo that. Which is a long winded way of saying that Claude is currently being used to identify bombing targets in Iran.
I'm not writing this to slam Anthropic. All the mainstream frontier models are involved in what I shall glibly call "shady shit". It's what pays the bills for phenomenally expensive operating costs.
This isn't just about AI - its a problem across all tech
I also don't think it's as simple as telling charities to "not use anything". Sure, it's a nice principle. But I think it's clear the train has left the station, and if your small charity isn't using AI in some capacity, sooner or later you're going to lose out to one that is. The point I'm trying to make is that if you are using AI, saying "well it's the good one" isn't going to work, because there are no "good" mainstream AI models.
But I'm also not having a go at the people making those decisions. It's fucking hard. There's a lot to consider and technology and it's allegiances change so quickly that it would be a full time job to keep up with just AI. A lot of these organisations don't have tech experts. They have a person who knows how to Google stuff better than their colleagues. (Not a dig. Do that for long enough and you get to call yourself an expert and charge people to tell them how to do their tech. Hi 👋)
Also, this isn't something that started with AI. The number of charities I see with Substack newsletters is staggering to me, given the platforms record of supporting and profiting from actual Nazis. But those organisations either a) don't know that (see paragraph above about how hard it is to keep up), or b) do know and they don't care. Or perhaps more accurately, they have made the determination that of the 87 million things they need to care about, the platform that welcomes Nazis is someone else's fight, and they just need to get their newsletter out quickly easily and cost effectively so they can continue to operate in a landscape that is becoming more and more hostile. Charities stayed on Twitter / X post-Musk-takeover for similar reasons.
Conclusion... sort of... not really...
My wider point, if I have one, is that you can't take any of these tech decisions at face value. People left OpenAI in droves because of a contract negotiation. And that's fine. But if your hard line is "I don't want to give my money to an AI company that is in Trump's back pocket", you'll be looking for a long time.
None of which is to say that you shouldn't have hard lines at all. You just need to be able to justify them if and when someone asks, and be willing to reconsider if you need to.
Want to chat about anything in this post? I'd love to hear from you!
Feel free to drop me an email
Or find me on Mastodon