AI did what? ChatGPT’s big upgrade, Google’s plans to make AI invisible, DeepSeek’s return and more of the week’s most surprising developments

• April 27, 2026

This week felt less like a collection of AI news stories and more like an early look at where the technology is really heading. Because yes, there are model upgrades and new features and disturbing stories this week, as there always are. But it all seemed to point to something broader, which is AI being positioned at the center of how we work, think, create and make decisions.

OpenAI is pushing ChatGPT toward an all-in-one “super app”. Google is turning AI into the layer that sits over everything you might do with your phone. And DeepSeek’s return is a reminder that AI isn’t just a Silicon Valley story anymore — it’s a global phenomenon. The AI race is becoming increasingly geopolitical.

I’ve also been thinking a lot about the bigger picture this week. Especially after Geoffrey Hinton, often called the “Godfather of AI” told a UN conference that rapid AI development needs urgent oversight. “If you ever went out with a car that had no brake, boy, you are in trouble if you go down a hill,” he said. “But you’re in even more trouble if there’s no steering wheel.”

Article continues below

I’ve also been looking into how AI tools are increasingly designed not just for your attention, but for your emotional attachment — and what that means for trust, influence and where people turn for support. None of that has easy answers, but I think it’s important to be asking the right questions right now as everything is moving so fast.

As always, I’ve pulled together the key stories you need to know below. Think you were paying attention this week? Take the quiz below to find out.

Were you paying attention last week? Take my AI news quiz

The top AI headlines from the past week

Welcome to ICYMI AI, your weekly round-up of the most important developments in artificial intelligence.Here are the biggest AI stories from last week and why they matter.

OpenAI is turning ChatGPT into “everything”

OpenAI released GPT-5.5 this week, framing it as a big step toward an AI “super app” that brings chat, coding tools and browser into one place.

It also launched ChatGPT Images 2.0, a major upgrade to its image generation. The focus is on better layouts, text within images and, maybe most importantly, reasoning, which means it can interpret more complex prompts. The output then reflects a sequence of decisions rather than a single pass.

The images upgrade is genuinely impressive. Well, until you try to do something specific with it. Our AI editor, a former print magazine editor, tested it on magazine layouts, and while the results look convincing at a glance, they fall apart the moment you need them to actually work. That gap between “looks real” and “is actually useful” is still very much there.

But the framing of ChatGPT as an AI super app is interesting. OpenAI doesn’t seem to be simply building a better chatbot anymore. It’s trying to build something closer to a whole ecosystem, where you write, search, build and make things. I guess whether that’s exciting or alarming depends on how much you trust OpenAI.

Google wants to change how you use apps

At its developer conference I/O in May Google is planning to announce that it’s making AI the main way people interact with their phones.

Android 17 and Gemini will handle everyday tasks automatically. You describe what you want, AI works out the apps to use without you opening them. The plan extends across devices, so something you start on your phone can be picked up on a laptop or in your car.

We’re still waiting on the official details from I/O, but this direction feels pretty significant. We’ve spent decades using smartphones by tapping an app, opening a menu, making a decision. Google might be changing how all of that works. You state the outcome, Google does the rest. That’s a fundamental shift in how we use technology. So, do we want a system that anticipates and completes actions on your behalf?

Chinese AI startup DeepSeek released preview versions of its new V4 model this week, claiming it’s the most capable open-source AI platform available with big improvements in reasoning and ‘agentic’ tasks. Unlike its previous models, V4 also runs on domestic Chinese chips from Huawei rather than Nvidia hardware.

A year on from the original DeepSeek news, the story has changed. Back then it was all about how much performance the company could get from relatively limited resources. Now, it’s about building a more self-sufficient AI stack that doesn’t depend on US technology in the same way.

That shift is important because it points to an AI landscape that’s fragmenting, with different regions building their own systems, infrastructure and standards. The 2026 AI Index from Stanford University found China leads in total AI research output and accounts for a large share of global patent filings.

The AI tools people use in the future may be shaped as much by geopolitics as by product design.

More AI news you might’ve missed

• Florida is investigating OpenAI: A criminal investigation has been launched into OpenAI after reviewing messages between ChatGPT and a student accused of a campus shooting. This week Sam Altman also sent a formal apology to the Canadian community of Tumbler Ridge where a shooter’s account was flagged and banned 8 months before a mass killing but not reported to police.
• Meta and Microsoft job cuts: Meta is cutting around 8000 jobs but ramping up AI investment. This pattern is getting really hard to ignore now — job cuts funding the tech that’s replacing them.
• Building a gaming PC with AI: We asked ChatGPT and Gemini to spec out a perfect gaming PC build. The results were enthusiastic but nearly cost us our sanity. A reminder that AI is confident, even when it probably shouldn’t be.
• Gemini notebook tips: Gemini’s new notebook feature is one of the more useful things to come out of Google’s AI push. Here are five ways to actually get value from Gemini notebook, from organizing research to summarizing long documents.
• AI personas are infiltrating online communities: A new study found AI fake personas are now realistic enough to join online communities and even steer public opinion. This raises big questions about whether you can trust anything you read online anymore.

Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.

This week felt less like a collection of AI news stories and more like an early look at where the technology is really heading. Because yes, there are model upgrades and new features and disturbing stories this week, as there always are. But it all seemed to point to something…