Trump's AI safety U-turn, Gemini Omni goes wild, and the 'AI washing' problem hits UK PR

Happy Memorial Day. Here's something that happened quietly on Thursday while most people were watching for AI news: a presidential executive order that was supposed to require safety reviews of new AI models before release got those provisions stripped out - hours before the signing ceremony.

That's not a minor edit. That's the whole point of a safety review policy, gone. And the timing tells you everything about who had the ear of the White House.

Let's get into it.

Cannes 2026 is featuring a major AI debate among filmmakers. But which pioneering AI film tool was famously used at a previous Cannes to generate controversy - earning its creators both acclaim and protests from artists?

The answer is hiding near the bottom of today's newsletter... keep scrolling. 👇

Here's what was supposed to happen: Trump would sign an executive order that included a government safety review process for new AI models before they could be deployed. Think of it as a checkpoint - a way for federal agencies to flag risks before the next generation of models hits the market.

What actually happened: just hours before the signing, that provision was quietly removed. The Guardian reports that the reversal came after significant pressure from big tech companies who argued the review process would slow down American AI competitiveness - particularly against China.

The argument the industry made was familiar: safety reviews mean delays, delays mean China wins the race. It's a frame that's proven remarkably effective in Washington, and it worked again here.

The result is an executive order that greenlights AI development with essentially no federal pre-deployment oversight. For labs racing to ship the next frontier model, that's a significant green light - and a significant loss for anyone who wanted structural guardrails built into the process from the start.

If you've been following our coverage of AI regulation, this fits a pattern: the US consistently choosing speed over structure.

Read the full story →

The Verge got hands-on with Google's new Gemini Omni model and the headline finding is this: it can transform anything into anything else. Video to audio. Images to video. Text to realistic scenes. The reporter used it to deepfake a stuffed animal onto a vacation - and it worked well enough to be genuinely unsettling.

The 'omni' framing isn't just marketing. The model handles multimodal inputs and outputs in a way that previous versions of Google Gemini couldn't. Earlier models could understand images or generate text - this one can take a photo of your dog and produce a short video clip of it running through a meadow.

The Verge's hands-on notes that it's 'almost there' in terms of quality - some generations still have the tell-tale weirdness of AI video - but the capability jump is real. For content creators, marketers, and honestly anyone with a camera, this is a meaningfully different kind of tool.

The bigger picture: as deepfake technology becomes this easy to use and this accessible, the 'almost there' part of that headline is doing a lot of work. It won't be 'almost' for long.

Read the full story →

UK public relations executives are sounding the alarm about what The Guardian is calling 'AI washing' - and honestly, the descriptions they're giving are pretty funny. Companies are doing 'yoga-level stretches,' according to one PR firm, to describe ordinary automation software as cutting-edge artificial intelligence.

The pattern goes like this: a company that has always used basic rule-based automation adds a chatbot interface, then wants their PR team to pitch them as an AI-native business. Or a firm that uses off-the-shelf analytics tools suddenly wants to be positioned as an AI innovator.

The PR executives interviewed say bosses are demanding they present these capabilities as AI in press materials - and they're frustrated because it makes their job harder when journalists push back on vague claims.

This matters beyond just honesty. When the term 'AI' becomes meaningless through overuse, it gets harder to communicate about genuinely significant developments - like the model releases and research breakthroughs that actually move the needle. It's the same problem that plagued 'blockchain' and 'big data' in their hype cycles.

If you're building something real with AI, this is a good reminder to be specific. What model? What capability? What measurable outcome? Vagueness is increasingly a red flag.

(Speaking of building something real - if you want to launch a legitimate AI-powered product fast, check out 60sec.site, which uses AI to build you a fully functional website in under a minute. No 'AI washing' here - it actually works.)

Read the full story →

The Cannes Film Festival has always been a place where creative ideology clashes with commercial reality - and this year, AI is the fault line. The Guardian reports on a split that goes right through the industry's most prestigious event.

On one side: Darren Aronofsky, speaking at an 'AI for Talent' summit on the Croisette, argued that AI tools 'expand the cinematic toolbox' and that the pushback from artists is understandable but ultimately misguided. On the other: Guillermo del Toro, who said he would 'rather die' than use AI in his filmmaking.

These aren't fringe positions - these are two of the most respected directors working today, and they represent a genuine split in how the creative world is processing this moment. Aronofsky's framing is collaborative and tool-focused. Del Toro's is about identity and what it means to make something.

The festival itself has become a proxy debate for the whole creative industries conversation: not just 'can AI make art?' but 'what happens to the people who made art before?'

Read the full story →

The Verge covered a story this week that the literary world is still processing: a short story published in Granta - as a regional winner of the Commonwealth Short Story Prize - appears to have been written by AI.

The piece, 'The Serpent in the Grove' by Jamir Nazir, shows what The Verge describes as 'many hallmarks of LLM-generated prose' - mixed metaphors, anaphora, lists of three, and a particular kind of sentence-level smoothness that human writers rarely produce consistently.

The story raises a harder question than just fraud. It's not simply that someone cheated - it's that the judges, editors, and peer reviewers at one of the most respected literary institutions in the world didn't catch it. The publishing industry genuinely doesn't have a reliable detection process yet.

And if it can happen at the Commonwealth Short Story Prize level, the question isn't whether it's happening elsewhere - it's how often.

Read the full story →

The answer is Runway Gen-2! The AI video generation tool made waves at Cannes in 2023 when a short film created using it screened at the festival, sparking both fascination and protests from artists and guild members who saw it as a direct threat to their livelihoods. It was one of the first high-profile moments where AI creativity collided with professional filmmaking at the industry's top table - and the debate clearly hasn't been resolved since.

Today's newsletter has a bit of a theme running through it - the gap between AI hype and AI reality. So here's my question: have you ever caught a company 'AI washing' - claiming AI capabilities that were clearly just basic automation? Hit reply and tell me the most egregious example you've seen. I read every response and the best ones might make it into a future edition!

That's all for today - enjoy the long weekend if you're in the US. Back tomorrow with more.

For daily AI news, tools, and breakdowns, visit Daily Inference.