The AI news cycle doesn't do slow weeks. I've been writing about it for more than a year now and this one still caught me off guard. Not because there's been one big breakthrough or story. But because I think there's been a mood shift. People are using AI more than ever, but a growing number seem to wish they didn't have to.
One of the stories I've been watching closely is Anthropic teasing its new model, Claude Mythos. It's described as a major step forward and yet the hype is running ahead of the evidence. Elsewhere, the focus is less on capability and more on broader consequences. Like Val Kilmer's return via AI, which raises questions about consent and the future of entertainment.
Once again, I’ve pulled together the key stories you need to know, plus a few practical prompts to get more out of tools like ChatGPT. Think you know the biggest AI stories from the past week? Take my AI news quiz below to see how much you remember.
Were you paying attention? 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.
The first full AI performance of a dead actor is here — and it won't be the last
Upcoming film As Deep as the Grave has used AI to recreate Val Kilmer in a major role. It's not a brief cameo, but a full performance with his family's approval.
This feels like a threshold moment. With AI, a person's likeness can become an asset that can outlive them. It does seem important to flag that the filmmakers are stressing Kilmer's family have been involved in the process. But that doesn't resolve some of the deeper questions around consent, ownership and what performance even means when an actor was never on set.
Another detail really stands out to me here. A key AI-generated shot reportedly only took minutes to make once the assets were ready. Which suggests this kind of AI movie-making isn't going to be a rare or expensive. We might get to a point soon where AI-generated movies are commonplace, but will audiences accept them?
- Read more: 'Don’t fear the dead, and don’t fear me' — AI brings a digital Val Kilmer back to the screen
A recent test found Google Gemini produces writing that's harder to flag as AI-generated than many of its rivals, particularly ChatGPT. Importantly, some AI detection tools failed to spot Gemini-generated writing at all.
This is significant because it undermines the AI detection tools many people are being told to rely on. Schools, publishers and some online platforms are investing in AI detection. But if results are inconsistent or just wrong, then the authority of those detectors becomes pretty meaningless.
The bigger issue is what happens when that boundary between human written and AI-generated content completely collapses. Plenty of people think they can still spot AI writing, but I'm starting to wonder whether we're only catching the most obvious tells and everything else is already slipping through unnoticed.
LinkedIn says AI isn't hitting jobs yet — but the evidence on the ground tells a different story
A LinkedIn executive said this week that AI isn't showing up as a reason for a drop in hiring. But stories from individual companies tell a different story. The latest example is UK supermarket chain Morrisons, which announced hundreds of office job cuts, explicitly tying them to AI restructuring.
This is part of a growing pattern across industries and countries. Targeted cuts linked to AI that don't seem to be registering yet in the big picture data. If your workplace has started talking about "efficiency" and "restructuring" in the same sentence as AI, you might already be seeing this play out. Even if that's not shown up in the hiring trends data.
Interestingly, it's not just jobs that might be shifting, the internet itself might be splitting in two. In an opinion piece, we explored the idea there could be soon a new internet, an 80/20 internet where 80% of web traffic is AI agents and only 20% is human. So there could be two internets, essentially, one optimized for machines and one for humans still seeking realness and authenticity.
More AI news you might've missed
- After a developer recently discovered that a popular plugin was collecting their AI prompts, it only feels right to share my favorite tips from the past week — this simple change in how you prompt AI and this prompt hack that fixes ChatGPT's biggest weakness.
- Love data and research? Me too. Stanford University has published its 2026 AI Index report tracking how AI's advancing fast but our ability to measure and manage it isn't keeping pace. And while we're on the subject of AI's fast growth, an ex-programmers speech at an Ohio planning meeting is going viral for highlighting what that growth actually costs on the ground – from drained water reservoirs to massive power demands.
- The Bank of England is running simulations on how AI could destabilize financial markets. Specifically, what happens if AI systems amplify a crash.
- Anthropic's new model, Claude Mythos, had its rollout restricted after it demonstrated the ability to find and exploit software vulnerabilities. It's worth getting clued up with this Claude Mythos explainer.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.