GPT-5.2 Rewrites Knowledge Work. Disney Licenses Characters for Sora. Google Opens Deep Research. FACTS Reveals Model Weaknesses.

Show notes

The AI news for December 12th, 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:

ChatGPT-5.2 is here: Why this update changes everything.
Source: https://www.all-ai.de/news/top-news24/chatgpt-5-2-release
Why did we choose this article?
Major model release with clear practical implications: big gains in accuracy, agentic tool use, long-context handling and reduced hallucinations make this immediately relevant for teams automating knowledge work and developers integrating smarter agents.

OpenAI partnership: Create your own Disney films with Sora starting in 2026.
Source: https://www.all-ai.de/news/top-news24/openai-disney
Why did we choose this article?
High-impact industry partnership that signals how major IP owners may choose licensing and product integration over litigation; practical for creators, platform engineers, and policy watchers planning for branded generative media.

Google introduces new deep-research agents and a new AI API.
Source: https://the-decoder.de/google-stellt-neuen-deep-research-agenten-und-neue-ki-api-vor/
Why did we choose this article?
Developer-facing update: Google upgrading Deep-Research with Gemini 3 Pro and an API that better supports agentic workflows matters for teams building autonomous tool chains and research assistants — signals where platform-level agent support is heading.

FACTS-Benchmark: Even top AI models struggle with the truth
Source: https://the-decoder.de/facts-benchmark-auch-top-ki-modelle-kaempfen-mit-der-wahrheit/
Why did we choose this article?
Critical reliability research: DeepMind's FACTS benchmark highlights persistent truthfulness gaps even in leading models — essential context for teams relying on LLM outputs and for designing verification layers and human-in-the-loop checks.

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