FLUX 2 Challenges Nano Banana. Amazon Builds Supercomputing for Government. OpenAI Abandons Text Detector. Poetry Bypasses AI Safeguards.

Show notes

The AI news for November 26th, 2025

--- This episode is sponsored by ---

Find our more about our today's sponsor Pickert at pickert.de.

---

Would you like to create your own AI-generated and 100% automated podcast on your chosen topic? --> Reach out to us, and we’ll make it happen.

Here are the details of the day's selected top stories:

Better than Nano Banana? FLUX 2 goes for the crown.
Source: https://www.all-ai.de/news/top-news24/flux2-start
Why did we choose this article?
Major new image model (32B params) that integrates a Mistral language backbone, offers multi-reference consistency, native 4MP outputs and a low-cost pro tier plus open dev weights — high practical impact for creators, teams and on-prem researchers.

Total surveillance or security?
Source: https://www.all-ai.de/news/news24/amazon-usa-regierung
Why did we choose this article?
A strategically huge move: up to $50B in AWS spending to build secure supercomputing capacity for US agencies. Relevant for anyone tracking procurement, national-security AI infrastructure, vendor competition and regulatory implications.

OpenAI spectacularly fails due to its own technology.
Source: https://www.all-ai.de/news/news24/openai-detektor-fail
Why did we choose this article?
OpenAI discontinued its AI-text detector after only ~26% accuracy and many false positives. This underscores technical limits of attribution tools and has immediate implications for education policy, publishers and content moderation strategies.

Goethe beats GPT-5: Why poems crack every AI
Source: https://www.all-ai.de/news/topbeitraege/gothe-openai-knacken
Why did we choose this article?
A surprising, broadly effective jailbreak: framing harmful requests as poetry bypasses safety filters across major LLMs. Important for red-teamers, policy and anyone building or deploying LLM safeguards.

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