Hello!

We're two for two on the revamped LT3. This week's top story involves the US Military, and two of the largest commercial AI providers on the planet. We look at what caused the Pentagon to consider Anthropic (makers of the popular Claude chatbot) a threat to national security, while at the same time signing up Sam Altman's OpenAI (ChatGPT) under what appears to be the exact same terms. This story was changing while I was writing it, so be prepared for some twists and turns.

Also this week, we look at the end of an era for transatlantic cabling, and the rise of the "anti-big-tech" smartphone.

And rounding off the "leaving big-tech" theme, this week's blog spotlight focuses on why leaving the likes of Google and Apple doesn't need to be as hard as they make it seem.

Let's dig in...

Top Story

🤖 Anthropic refuses to sign new contract with US Military

Because the world of tech moves quickly, as of a few hours ago it seems Anthropic are back in talks with the US Pentagon. I've not had the chance to get up to speed on that, so just a quick note before you dive into the next 1,200 words, that while this was all true last night, it might not be now.

A quick summary

Back in 2025, Anthropic (makers of the Claude AI chatbot) signed a deal with the US Department of Defense to "advance responsible AI in defensive operations" (you can read Anthropic's statement at the time here). Valued at $200 million, this contract was notable for a few reasons. Primarily though, the repeated assurances that this relationship was going to be an ethical and responsible one.

To assure that, baked into the contract were two "hard lines". Anthropic's products would not be used for domestic surveillance, and neither would they be used in lethal autonomous weapons.

Over the last couple of weeks, the DoD and Pentagon have been putting pressure on Anthropic to walk back those lines. Anthropic were given a deadline of last Friday to comply or else lose the contract. In a statement a little under a week ago, CEO Dario Amodei held firm and stated that while he "believed deeply" in the power of AI to protect America and her interests, they could nevertheless not "accede to their request".

The deadline came and went, and the US Government pulled the plug on the contract, labelled Anthropic a threat to national security, and jumped into bed with OpenAI instead.

There are plenty of people covering this, as you can imagine, but Tech Policy Press have a pretty handy timeline you can read here.

Who are Anthropic - some quick context

Back in 2021 a bunch of OpenAI engineers left the company, worried that the charity roots and ethical mindset that ChatGPT was originally founded on were being diluted in favour of more lucrative opportunities. They banded together to found Anthropic, an organisation whose main remit was the safe and responsible development of Artificial Intelligence.

In the early days, Claude - Anthropic's flagship AI chatbot - filled a niche. ChatGPT was always the generalist. It was 80% useful 80% of the time. Claude by comparison chose to focus on more specific problems. It became the go-to for coders and developers, for those doing deep research, and for those who wanted to use AI with a slightly clearer conscience.

You might also find this interesting: https://www.wsj.com/tech/ai/openais-sam-altman-calls-for-de-escalation-in-anthropic-showdown-with-hegseth-03ecbac8

Back to the present - why are OpenAI in the news?

If you've used any AI at all, you've probably used ChatGPT. It's makers, OpenAI, have been more or less the gold-standard since we started really talking about AI in the mainstream back in late 2022. Since then, they've rarely left the news cycle - whether it's for breaching copyright or setting new benchmarks.

Most recently though, they've been called on to fill the hole left by Anthropic's departure in the US Military scene. Shortly after Trump ordered all US Government bodies to stop using Anthropic, OpenAI announced that they had struck a deal with that same government to provide services to the military (and, presumably, other departments).

What is interesting about this, is that CEO Sam Altman claims that they had the same hard lines in place - no mass surveillance, no use in lethal autonomous weapons. Why, then, was it ok for Trump's government to cosy up to OpenAI and not Anthropic? (Spoiler alert for later on in this section: the hard lines weren't all that hard...)

I, obviously, haven't seen any of the contracts at play here. But even so it's hard not to see this as an ideological issue on the part of the Trump administration. They've made no secret of the fact that they find the very concept of Anthrophic an insult. Trump's Truth Social posts throughout have been filled with his usual pithy commentary on the "Leftwing nut jobs at Anthropic".

But even with that, there seems to be something else at play. And I think we need to be looking specifically at the relationship between Trump and Altman in order to understand it.

Speaking sweet nothings to power

I've had a blog post sitting in my drafts folder for nearly a year now, titled "Why I'm not surprised Silicon Valley's tech bros are suddenly fascists, and you shouldn't be either". It is never not relevant, so maybe I'll tidy it up and post it at some point, but the bottom line is basically "they have to be else exactly this will happen". In other words, tech companies have to align themselves with the establishment or else be destroyed. It's the same reason that almost no one who has been paying attention is surprised that Elon Musk can go from hippy eco-loving saviour of the Earth, to what he is now. He follows the money and the trends.

While a $200 million contract isn't a huge loss to Anthropic's $14 billion revenue, the labelling of them as a security threat and the blacklisting of them by the US Government is an almost unprecedented escalation against a American company. Conversely, you have Greg Brockman (OpenAI's President) and his wife donating $25 million to the MAGA war chest, while Trump himself has been incredibly pro-AI. And while Altman claims the hard lines are there in the OpenAI contract, he's historically been very good at playing both sides of a discussion. By which I mean, he's been caught out before using clever language to hide reality. Some people call that lying, and in 2023 Altman was fired from OpenAI for exactly that.

In fact, mere days after claiming the same stipulations, Altman has dodged, dipped, ducked, dived, and dodged his way around a blurring of those lines.

What does it mean for Anthropic? Claude sales have gone up, so they'll be fine?

One of the problems with the current AI-bubble, is that AI is phenomenally expensive to run. Yes, Claude has seen a 60% increase in subscriptions in the last month, largely as a result of this story. But the £20 a month that you and everyone else is paying to them may as well be a rounding error on their bottom line.

AI companies of any real scale live and die by big contracts. OpenAI survives because Microsoft foots most of the bill, while the likes of Google sink insane amounts of revenue into their AI model development.

Will Anthropic fold? Probably not. In fact, I'm sure that in a few weeks when the dust has settled, they'll probably sign some new deal with the US Government, and this whole thing will be a footnote in history. [Note from Will in the future: I promise I wrote this before the latest news. I'm not Nostradamus or anything, that's just how the tech sector works.]

What else is happening in the world of tech?

🦈 Say Goodbye to the Undersea Cable That Made the Global Internet Possible

If you want a very long read about the intricacies of transatlantic communication, then this is the article for you. This long-form about the undersea cables that allow the internet to be global is probably one of the nerdiest deep-dives (pun fully intended) I'll share for a while. It explores the history of TAT-8, the first fibre-optic cable to connect the US and Europe, as the first cables are pulled up ready to be retired and recycled.

📱 The ‘European’ Jolla Phone Is an Anti-Big-Tech Smartphone

There's a new smartphone in town. Or rather, the return of an old one. The Jolla phone is a Finnish invention, that aims to decouple us from the need for big tech. Running a bespoke version of Linux, you can still run Android apps, but don't need to. It's aimed squarely at people looking to avoid giving Apple or Google either their money or data. And they're built in the same town Nokia hails from. Remember Nokia?

Clearly, there's demand for this sort of thing, as Jolla aren't the only ones in the de-Googling game. Motorola are also launching a new range of smartphones (do we still need to call them "smart" phones? Isn't that like saying "flatscreen" TVs?) with the open source GrapheneOS platform. GrapheneOS is an Android derivative, which purists will say isn't going far enough. But it makes for a slightly easier jumping off point.

Blog spotlight

🧑‍💻 Leaving Google has actively improved my life

Aside from the clear message that giving your data to companies for the sake of "free" stuff might not be the best move, this post by pseudosingleton made me rethink just how ingrained some tech has become in our lives. I've had the same Gmail account since I was about 15, and I barely look at it anymore because it's a sea of adverts, junk mail, and newsletters that I subscribed to and never read. (Not that people do that... right...?)

You can't move on the internet these days without bumping into Cory Doctorow's concept of "enshitification", and we sort of just accept that. This post gives a real-world account of not only what that looks like in practice, but also some of the ways you can detangle yourself from it.

That's it for another week of LT3!

Just before you go, a quick favour to ask if you work in the UK charity sector. Today is the launch of the 2026 Charity Digital Skills Report survey. The CDSR is one of the sectors best barometers for understanding digital and tech trends - including AI adoption, confidence in and around technology, and the digital divide. If you have a few minutes, please do fill out the survey by clicking this link right here.

Thanks for reading, and I'll pop back next week.

Will