Coding Assistants Hijacked. Extensions Leak AI Chats. ACCA Ends Online Exams. Laws of Reasoning Improve AI.

Show notes

The AI news for December 30th, 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:

39C3: Security researchers hijack AI coding assistants with prompt injection
Source: https://www.heise.de/news/39C3-Sicherheitsforscher-kapert-KI-Coding-Assistenten-mit-Prompt-Injection-11125630.html?wt_mc=rss.red.ho.themen.k%C3%BCnstliche+intelligenz.beitrag.beitrag
Why did we choose this article?
Concrete security research demonstrating prompt-injection attacks against coding assistants. Important for practitioners and teams relying on AI coding tools — shows what kinds of vulnerabilities remain, which fixes help, and why defense-in-depth and prompt/sandboxing practices are necessary.

Security researchers warn about browser extensions that secretly intercept AI chats.
Source: https://the-decoder.de/sicherheitsforscher-warnen-vor-browser-erweiterungen-die-heimlich-ki-chats-abgreifen/
Why did we choose this article?
High-priority privacy risk: extensions can exfiltrate sensitive chat data from web-based AI tools. Actionable takeaway for users and orgs — audit and restrict extensions, treat chat transcripts as high-risk data, and prefer vetted clients or policies that limit extension access.

Due to AI fraud: the world's largest accounting organization abolishes online exams.
Source: https://the-decoder.de/wegen-ki-betrug-weltgroesste-buchhaltungsorganisation-schafft-online-pruefungen-ab/
Why did we choose this article?
Shows a tangible institutional response to widespread AI-assisted cheating: removing online exams. This has broad implications for credentialing, remote assessment design, proctoring tech, and how organizations balance accessibility with trust.

A team of researchers wants to end illogical AI babble with new 'Laws of Reasoning.'
Source: https://the-decoder.de/forscherteam-will-unlogische-ki-gruebelei-mit-neuen-laws-of-reasoning-beenden/
Why did we choose this article?
Research relevant to core model behavior: proposes formal 'laws' to fix illogical or inefficient reasoning in current models. Valuable for developers and product teams tracking where model capabilities are headed and what that means for reliability and safety of reasoning-heavy applications.

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