Paris Hilton Takes On AI Deepfakes

AI Around the World

US: Paris Hilton vs. Deepfakes Paris Hilton warned lawmakers that AI sexual imagery is the “newest form of victimization.” After being targeted by 100,000+ deepfakes herself, she’s backing the DEFIANCE Act - a bill that empowers victims to sue creators.

Finland: The Talent Poach Finland is aggressively recruiting US AI talent with a simple pitch: Swap Silicon Valley burnout for Nordic stability. The carrot? A fast-track work visa approved in just two weeks.

China: The First Humanoid Astronaut? Engine AI wants to send its robot, PM01, to orbit. The mission is to test if a humanoid can survive the vacuum and radiation of space - handling the risks so humans don’t have to.

Saudi Arabia: AI for the Superfan Saudi Research and Media Group just launched GOAT, a new sports app. It uses AI to cut through the noise - curating real-time scores and highlights, with a laser focus on the Saudi Pro League.

Austria: The Viral AI Assistant Austrian developer Peter Steinberger accidentally broke the internet with Clawdbot (now Moltbot), a self-hosted "AI employee" that runs 24/7 on your local machine. The viral hit caused a sudden run on Mac Minis as techies rushed to build their own "Jarvis" … ignoring warnings that giving an autonomous AI root access to your files is a "security nightmare."

Time to Stop Being Nice to ChatGPT?

Most of us default to being polite with machines. We say please. We say thank you. We treat chatbots the way we were taught to treat people.

But a new study found something surprising: when researchers used ruder, more demanding prompts, ChatGPT’s answers actually became more accurate. Accuracy jumped from about 76% with very polite prompts to nearly 85% with very rude ones. 🤬

It doesn’t mean AI has feelings - it’s just that harsh prompts tend to strip away any vague language, forcing the system to respond more directly instead of offering safe, generic replies.

Tom Lue’s Mission to Govern AI

Powerful AI. Weak rules. That mismatch won’t end well.

That’s why Tom Lue’s role at Google DeepMind matters. As VP of Frontier AI Global Affairs, he works on governance before systems scale, not after trust is lost.

Tom is uniquely suited to this. With a background spanning the White House, Department of Justice, and Waymo, he bridges the gap between "Frontier AI" (highly capable systems with novel risks) and the laws that guide them.

But this isn't just about avoiding disaster; it's about unlocking potential with responsible AI.

In this episode, Tom shares a stunning example of what happens when we get it right: DeepMind’s AI models improved monsoon predictions so significantly that they helped double the annual income of farmers in the India.

It’s proof that the goal isn't slowing AI down … and we can steer it toward human benefit.

Don't miss this fascinating conversation: YouTube or Spotify.