You’re receiving this email because you subscribed to the newsletter. If you’d like to unsubscribe, you can manage your preferences here with one click (or use the button at the bottom of the email).

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 ‏‏‎ ‎

Let's jump in!

1. X allows its AI to harm innocent people

We start the year with quite a depressing, but incredibly important story. Over the last few days, users on X have been using the platform’s chatbot and asking it to “undress” images of women. In some cases, they even asked the chatbot to do this for children.

Regulators from the UK, EU, Australia, Brazil, and India have all raised concerns about this, but no firm action has been taken yet by governments - which is quite shocking, given the seriousness.

Following days of pressure, Elon Musk’s team has now banned free users from doing this - but paid users are still allowed. The UK Government has said this change is “insulting to victims” and will back the local regulator if they decide to ban X.

To be clear, this is something that X and Elon Musk have allowed on the platform. They do have the ability to stop this from happening and can easily stop their AI model from generating these appalling images, but they’ve decided not to.

For all the incredible things that AI models can do, this shows just how easily they can be used to harm innocent people. It’s a reminder that we need strong regulations and for governments to take action - because clearly, some platforms cannot be trusted to police themselves.

2. Google lets teachers convert their lessons into podcasts

In more positive news, Google has released a new tool that allows teachers to easily generate audio lessons (like a podcast) and share them with their students.

To do this, you simply include the grade level, topics you want to cover, learning objectives, and then customise the format - which could be an interview, roundtable discussion, or a more casual conversation between several people.

Google’s AI model will then generate the audio for you, check that you’re happy with the result, and share it with the students. Overall, it seems like a very good way to interact with students - especially those who are auditory learners and remember things by how they sound.

In recent years, there has been a shift in how people consume media - with more people turning to podcasts for information instead. In 2023, 24 million Gen Z listeners in the US were tuning into podcasts every month. By 2025, that number jumped to over 35 million people.

Google’s clearly responding to this global trend, giving teachers a simple way to tap into it and reach students in a format that works for them.

3. China pushes for stricter chatbot laws

China has unveiled draft regulations that would impose strict safeguards on artificial intelligence, with a particular focus on protecting children and preventing chatbots from encouraging harm or violence.

The rules, which were published by the Cyberspace Administration of China, would require AI developers to include specific settings for younger users, enforce time limits, and ask for a guardian’s consent.

Notably, chatbot companies must have humans that can intervene in conversations - including those that mention depressive thoughts or harm - and can immediately alert the person’s emergency contact.

It follows a huge rise in advanced chatbots globally, with tens of millions of Chinese now using these tools. In many cases, people are turning to chatbots for companionship and even therapy, raising questions about how the technology could affect our mental health.

4. Samsung says its profits will triple, as memory prices increase significantly

Bad news for ordinary consumers, as memory prices continue to skyrocket globally and are projected to do the same in 2026. We could start to see these prices filter down into phones, tablets, and laptops in the coming 12 months - as manufacturers start to grapple with significantly higher costs.

But one company that’s doing incredibly well is Samsung. The company is forecasting that operating profits will triple to 20 trillion won and is on track to reach record profits, with much of this being fuelled by the huge demand for AI components.

As companies scramble for memory chips to power their AI processors, that demand is now being felt across the market and is squeezing supply. Memory prices have climbed 40-50% in Q4 alone, with similar increases expected through early 2026.

5. AWS CEO says replacing juniors with AI is "one of the dumbest ideas” ever

Since ChatGPT landed in late 2022, companies haven't stopped talking about AI and its impact on jobs. We’ve also seen a dramatic slowdown in the hiring of junior employees - Big Tech brought on 25% fewer graduates in 2024 compared to the year before.

Some executives have openly discussed replacing entry-level staff with AI, which is "one of the dumbest ideas" that AWS’ CEO has ever heard.

Matt Garman argues that this strategy makes little sense and will only harm companies in the longer-term. Junior employees are often the most adept at using AI tools, as they’ve grown up with the technology.

They're also the cheapest to employ, so the cost savings are quite small. Plus, eliminating these roles destroys the talent pipeline that companies need and puts their long-term growth at risk.

That said, Garman isn't dismissing AI's impact entirely. He's been pretty direct with Amazon staff that their jobs will change - as the skills that made someone successful five years ago won't be the same ones needed next year.

It’s a refreshing and measured take on the technology, which has attracted huge amounts of hype for the last three years. If companies think they can replace their junior employees and skip hiring altogether, they're setting themselves up to fail.

Meta’s new audio AI is a goldmine for advertisers

There was an interesting announcement from Meta, who have been able to create an AI model that can isolate specific sounds in an audio file.

For example, let’s say that you’ve recorded a video and there’s a loud train sound in the background. With this AI model, you can click on the train and remove that specific sound - allowing you to hear the other parts of the clip - like birds singing in the distance.

Or if you want to focus on that sound and remove everything else, you can do that too. You can also type specific sounds, like “dog barking”, and it will filter that sound for you.

It’s an incredibly clever tool, which uses a Perception Encoder Audiovisual to align what's seen with what's heard. It’s critical that this works well, as the system needs to understand that the person moving their lips is the one who’s speaking - not someone who’s off-camera.

As a result, Meta had to train the system on a huge number of videos - with over 100 million used and this would have been incredibly expensive to produce. But there’s a clear business case for Meta here and they’ll make that money back several-fold.

Since advertising is such a huge proportion of their revenues, they can use SAM Audio to translate an advertiser’s video into other languages. This tool can identify who’s talking, focus on that specific part of the audio, remove it, and then seamlessly add the translation.

These kinds of features are what make Meta’s platform so powerful. It’s got one of the best recommendation engines out there and this will enable advertisers to get even better results. If you want to try SAM Audio for yourself, I’ve included a link to it below.

Google allows AI agents to create user interfaces

If you’re developing AI agents, this is something you’ll want to look at adopting. Google has recently announced A2UI, which is their new standard for interacting with AI agents and allowing them to generate user interfaces.

It essentially provides AI agents with a common language for describing interfaces, without the security headaches of running untrusted code. This standard also allows companies to add their own branding and make the chat interface look more like their own app, rather than a simple chatbot.

For example, a restaurant booking agent can use it to generate a date and time selector - similar to how you would in a normal app - and ensures that users don’t have to constantly respond with text messages.

Several Google teams have already adopted A2UI internally, including Gemini Enterprise and their experimental Opal mini-apps platform. CopilotKit's AG-UI framework also supports the standard, which will allow more developers to quickly get started.

I’ve been using A2UI recently for my own professional work and it does work reasonably well. But there’s still more work to be done on simplifying the set-up and allowing us more options for customising the UI.

It’s still a great first step and I expect that this will be a growing trend in 2026. After three years of typing into chatbots, we need to provide users with a richer and more familiar brand experience - which is now possible thanks to A2UI.

If your teams haven’t started looking into this area yet, which I highly recommend you do, I’ve included a link to the documentation below.

🏥 OpenAI releases ChatGPT Health, 230 million people ask about health each week

🚁 UK invests £140m ($190m) in drone technology to counter Russia

🛒 Shopify competitor Swap raises $100 million

🚜 Caterpillar will integrate AI to its construction equipment, partners with Nvidia

🔏 Meta releases an open-source watermarking tool for AI

💼 OpenAI acquires an executive coaching AI tool

🤖 Ukrainian robot has successfully fought Russian forces for 45 days

🏭 Boston Dynamics will use robots to build Hyundai cars in 2028, partners with Google DeepMind

🚗 Nvidia's AI model allows autonomous vehicles to "think like a human"

📉 Tesla's annual sales decline 9%, BYD overtakes it as the global EV leader

🇪🇺 EU waters down its 2035 electric vehicle plans

🎬 Luma's new AI model can generate videos using a start and end frame

For the last year, I’ve been closely watching CFS and the advances they’re starting to make with fusion power. If you haven’t heard of them before, they’re an American company based in Cambridge, Massachusetts and spun out of MIT in 2018.

The startup aims to build a small power plant, capable of generating fusion power, which could lead to almost limitless clean energy. It’s just reached a significant milestone and several superconducting magnets are now installed within its reactor.

Each magnet weighs 24 tonnes and generates a 20 tesla magnetic field, which is strong enough to lift an aircraft carrier. The company expects all 18 magnets installed by summer's end, with the reactor switching on next year.

CFS has also partnered with Nvidia and Siemens to build a digital twin of the reactor, allowing engineers to run experiments virtually before applying changes to the physical system.

Overall, the company has raised a whopping $3 billion in funding - with backers including Google and Nvidia. For tech companies, the benefits are quite straightforward.

If CFS can reach a breakthrough and start generating fusion power, they would have limitless amounts of power for data centres and future AI systems - alleviating a huge bottleneck.

Microsoft has taken similar steps and is investing significantly in Helion Energy, which aims to bring its reactor online by 2028.

Interestingly, President Trump’s company has also made moves in this sector. In December, his social media arm announced that it would merge with TAE technologies, with the merger valued at $6 billion.

We still need to wait and see how this plays out. There are more hurdles for CFS and Helion to overcome, but both companies aim to properly test their technology in the next 2 years. In the meantime, investors are scrambling to get in on the action.

This Week’s Art

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

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

Or if you’re interested in chatting with me about the above, simply reply to this email and I’ll get back to you.

Liam

Share with Others

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

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.