EPA Power Plant Loophole. Perplexity Brings AI to Windows. Hugging Face Hosts Nonconsensual Deepfakes. AI Finds Crypto Flaws.

Show notes

The AI news for July 29th, 2026

--- This episode is sponsored by ---

Rocket Routine GmbH
Find our more about our today's sponsor Rocket Routine at
SO2: No U.S. emission limits for fossil power plants that only supply data centers.
Source: https://www.heise.de/news/Saurer-Regen-Emissionsgrenzen-gelten-nicht-fuer-US-Kraftwerke-fuer-KI-11381087.html?wt_mc=rss.red.ho.themen.k%C3%BCnstliche+intelligenz.beitrag.beitrag
Why did we choose this article?
US regulators exempting off-grid power plants from emissions rules directly enables data-center operators to run fossil generators without the usual limits — a concrete environmental and community impact tied to AI infrastructure expansion.

Perplexity’s Personal Computer turns Windows PCs into AI agents
Source: https://www.theverge.com/ai-artificial-intelligence/971750/perplexity-personal-computer-windows-ai-agents
Why did we choose this article?
Windows users can now run a local AI 'agent' that accesses files and apps — a direct change to personal and workplace workflows with immediate privacy, security, and productivity implications.

Hugging Face is being used to easily undress women and children
Source: https://www.theverge.com/ai-artificial-intelligence/971723/hugging-face-nudify-deepfake-undress-women-children
Why did we choose this article?
The finding shows mainstream model hosting enables easy nonconsensual deepfakes of vulnerable people — a direct user-safety and moderation problem for platforms, creators, and victims.

Anthropics AI model Mythos Preview finds vulnerabilities in cryptographic algorithms that secure the Internet.
Source: https://the-decoder.de/anthropics-ki-modell-mythos-preview-findet-schwachstellen-in-kryptografischen-algorithmen-die-das-internet-absichern/
Why did we choose this article?
An AI model was able to discover weaknesses in cryptographic algorithms — a warning for security teams and infrastructure owners that AI can probe and potentially undermine core internet security assumptions, even if no immediate breaches were reported.

Do you have any questions, comments, or suggestions for improvement? We welcome your feedback at podcast@pickert.de.