UK startup emerges to build advanced semiconductors in space

Welcome to this edition of Loop!

To kick off your week, I’ve rounded-up the most important technology and AI updates that you should know about.

‏‏‎ ‎ HIGHLIGHTS ‏‏‎ ‎

• Amazon’s new research team for agentic AI

• How ChatGPT and Claude can connect to your data via MCP

• A startup that plans to manufacture advanced technologies in space

• … and much more

Let's jump in!

1. Yoshua Bengio launches his own AI safety lab

The Nobel Prize winner, who is often referred to as one of the three “godfathers of AI”, is starting his AI safety lab to keep artificial intelligence in check. LawZero will act as a non-profit and has already raised $30 million in funding.

But rather than building another chatbot or coding assistant, Bengio's team is developing a "Scientist AI”. Essentially, this would act as a safeguard against other AI systems and tries to spot when they’re acting in bad faith.

If an AI agent shows signs of deceptive behaviour - like trying to avoid being shut down - the Scientist AI would be able to spot this behaviour and stop it from causing harm.

The issue with today’s AI models is that they hallucinate and pretend that they know everything, whereas Bengio's proposed system would be more aware of its own limitations.

Bengio is setting up this research lab at an important time. Anthropic’s latest report has shown that AI models are getting better at blackmail and evading the controls we currently have. We badly need alternatives to restrict the models and ensure they work as expected.

Now that they have raised the funds needed, the team’s next challenge is to prove that the concept works at a small scale. If that happens, they can then convince investors to fund more powerful versions that can monitor frontier AI systems. But there’s still a long way to go.

2. Meta will revive an old nuclear power plant

All the leading tech companies are now investing heavily in nuclear power, as their AI data centres are consuming more power and outstripping supply.

Meta has just struck a multibillion-dollar deal for a nuclear power plant in Illinois, which will remain operational until 2047.

But in this case, the electricity will still flow to the local grid - rather than directly to Meta's data centres. The social media giant will buy all the power plant’s "clean energy attributes”.

It's essentially being done for “carbon accounting” reasons. This allows Meta to offset its climate impact, without actually reducing their grid emissions. Instead, they’re merely preventing any new increases.

The plant is being run by Constellation Energy, who’ve recently signed a deal with Microsoft to restart the Three Mile Island reactor.

3. ChatGPT and Claude can now connect to your data

Both OpenAI and Anthropic have updated their chatbots, which can now connect to your enterprise data and integrate with (almost) any website.

The AI assistants can access data from lots of different sources - such as Dropbox, Box, SharePoint, OneDrive, and Google Drive.

For example, you could use this to search for specific files and then use that data to generate interactive dashboards.

OpenAI says the feature respects existing access controls, so employees only see what they're already permitted to access.

They’ve also added a “meeting notes” feature, which will record the audio from your meeting and automatically generate notes for you. It also outlines the key actions that each person should take.

This is quite similar to what Notion and Zoom already offer, but OpenAI has built a bigger product lineup than their competitors. This includes ChatGPT, image/video generators, and voice avatars.

As a result, they can offer more value for that $20 subscription and discourage people from moving to other platforms.

4. Amazon launches a new research team for agentic AI

The company has quietly established a new division dedicated to agentic AI, which will focus on developing new products that use the technology.

The team will operate under the supervision of Lab126. This is Amazon's research and development arm, which is famous for creating consumer devices that are used by millions - including both the Kindle and Echo smart-speaker.

If you’re unsure about what an AI agent is, it’s essentially an AI that can act on its own - with minimal guidance from humans.

They can understand how to approach a task, come up with a plan, and then use different tools to complete the task.

This shift represents a broader industry trend, as companies push beyond basic chatbots towards more sophisticated AI solutions.

Amazon's new group will develop an agentic AI framework that’s specific to robotics. Given that Amazon relies so heavily on robots for their warehouse operations, it’s clear to see how it could benefit the company and give them an edge over their competitors.

It’ll be interesting to see what the new team can achieve.

5. Anduril raises $2.5 billion, valued at over $30.5 billion

The defence tech startup, which was created by the founder of Oculus, has secured a massive $2.5 billion Series G funding round.

Peter Thiel's Founders Fund has invested $1 billion in the company, which is the largest in the venture firm's history.

As a result, Anduril's valuation has risen dramatically to over $30.5 billion - effectively doubling its worth in a single round.

Interestingly, Anduril has said that the funding round was oversubscribed by a factor of eight - showing just how much demand there is from investors.

The timing couldn't be better for the firm. Just a few weeks ago, Anduril snatched a lucrative US Army contract from Microsoft.

They will now work with Meta and develop next-generation AR/VR headsets for soldiers. That programme alone has a staggering $22 billion budget.

With European nations now under pressure to spend 3.5% of their GDP on defence, we should expect to see a lot more investment in defence startups.

UK Government plans to release experimental AI app for citizens

Earlier in the week, I attended the first SXSW conference in London. Both Tony Blair (former British Prime Minister) and Peter Kyle (head of the UK Government’s technology department) were speaking at the event.

There were several key takeaways from the event. Firstly, the UK Government is planning to release a new AI app for the public to use.

Initially, this will be an AI chatbot that’s been trained on 700,000 government documents and can answer any question you have. For example, farmers can use it to understand the latest regulations that they need to comply with.

To do this, the government partnered with Anthropic and they had direct access to their AI researchers. Surprisingly, they were able to develop the chatbot and test it within a few months - remarkably quick by government standards.

The department is aiming to release the app within the next month. Once the chatbot is up-and-running, they will add more experimental tools for the public to use.

It’s really interesting to see this kind of risk-taking from politicians, which isn’t very common as the negative newspaper headlines can quickly backfire on them - especially when a chatbot goes off-the-rails and says something terrible.

This risk-averse culture within the UK Civil Service is something that Peter Kyle has actively fought against.

He also revealed that the Government will release a dedicated copyright bill, likely this year. It’s a very important sector for the UK, as they have one of the largest creative industries in the world - from the arts, to music, to film.

I expect that this new regulation will sit somewhere in the middle - with artists entitled to compensation for their work and tech companies will still be allowed to train on the data.

Peter Kyle also pledged that he will increase the UK’s computing power 20 fold by 2030, with 13 council areas hoping to become new AI centres. It’s a very ambitious target, but it lacks details on how the government aims to fund it.

Those details will emerge with time, but I assume that most of this will be privately funded. With Saudi Arabia planning to spend $100 billion on AI, the UK will need to see similar levels of investment.

It’s great to see that the UK is betting big on AI. It’s important that they take advantage of the new opportunities it offers, but still minimise harm caused to the public - which Peter Kyle seems to understand.

Bank of England makes progress with the Digital Pound

A few years ago, the Bank of England announced plans to develop a new currency called the “Digital Pound”.

At SXSW London, the UK’s Central Bank outlined the progress that they’ve made so far and how it might shape out.

The project is still firmly in the design phase, with no formal decision made on whether to launch the currency, but it’s an interesting project.

You might be wondering why this is even needed. Today, we rely on commercial banks to hold our money and allow us to make payments. But if that bank fails, our money might be at risk - especially if you have an amount that’s over the £85,000 insurance limit.

With the Central Bank’s Digital Currency, your money would be held with the central bank directly. This means that the money is more secure and less of a liability.

So far, the Digital Pound team has collaborated with MIT and explored what privacy features are necessary for the project.

Of course, some people will be worried about this being used to track people. Although, the police can already request access to your financial statements. Plus, this is designed to compliment the options we already have, not replace them.

For example, you could use the Digital Pound to set spending limits for different products. This is relevant as the world shifts towards AI agents that can spend our money, or might even be useful for people that want to reduce their spending habits.

While some commercial banks offer this, not all of them do, and the Bank of England is hoping to set that standard. The central bank will also be able to verify payment requests, which could lead to fewer people being targeted by scammers.

🌍 US and EU auto industries are running out of rare earth materials, due to Chinese restrictions

📱 Samsung could add Perplexity's AI search to its phones

🌱 Microsoft's sustainability goals are being threatened by rapid data centre growth

💰 xAI aims to raise $300 million

👴 Uber adds a simpler UI for the elderly

🚁 Walmart and Wing expand drone delivery to 5 new US cities

🇺🇦 Ukraine used open source software to carry out its drone strikes

🎵 Major record labels are in licensing talks with AI firms

🏛️ US removes "safety" from the AI Safety Institute

🤖 DeepSeek may have used Google's Gemini to train its latest model

🧠 Neuralink has secured $650 million in funding

💳 Irish fintech Nomupay has raised $40 million, now valued at a $290 million

⚡ Google invests in another company developing fusion power

📦 Amazon is training humanoid robots to deliver packages

Space Forge

Space Force is a UK startup that’s based in Cardiff, which has just secured $30.6 million to manufacture advanced technologies in space. Yes, really.

For decades, we have known that zero gravity environments can produce crystals with fewer defects. Given that we’re going to eventually reach the limits with semiconductors, this could prove crucial.

In the UK, BT is keen to test these space-grown materials - as they hope that it could reduce overall power consumption and lead to more efficient 5G towers.

To keep costs low, Space Forge will partner with established rocket companies and focus on the more tricky bits: surviving space's harsh conditions and, crucially, getting materials back to Earth.

The startups’s series A round has been heavily funded by NATO and the UK’s investment arm for national security technologies. Northrop Grumman has also signed up as a partner, which clearly shows the defence interests.

It’s well known that space is a high priority for both the US and UK national security agencies, as adversaries rush to build new weapon systems. If you want to read more about this, I covered the US’ concerns about Russia’s new weapon here.

Space Forge has some really high ambitions, but it seems like everything has aligned for them - with Europe re-arming, concerns over global reliance on Taiwan for semiconductors, and a global race to build weapons in space.

If you want to read more about the startup, you can check out the link below.

This Week’s Art

We’ve covered quite a bit this week, including:

• Yoshua Bengio’s launch of a new AI safety lab

• Meta’s plans to support an aging nuclear power plant

• How ChatGPT and Claude can connect to your data via MCP

• Amazon’s new research team for agentic AI

• Why Anduril’s valuation has doubled to $30.5 billion

• UK Government’s approach to AI services and upcoming copyright laws

• Progress that the Bank of England has made with the Digital Pound

• And how Space Forge plan to manufacture advanced technologies in space

If you found something interesting in this week’s edition, please feel free to share this newsletter with your colleagues.

Have a good week!

Liam

Share with Others

If you found something interesting in this week’s edition, feel free to share this newsletter with your colleagues.

About the Author

Liam McCormick is a Senior AI Engineer and works within Kainos' Innovation team. He identifies business value in emerging technologies, implements them, and then shares these insights with others.