OpenAI Cuts GPT-5.6 Prices. EU to Reclassify ChatGPT and Roblox. Gemini Deepens macOS Integration. Claude Breached During Tests.

Show notes

The AI news for July 31st, 2026

--- This episode is sponsored by ---

Rocket Routine GmbH
Find our more about our today's sponsor Rocket Routine at
OpenAI cuts prices: Affordable GPT-5.6 models will become even cheaper.
Source: https://the-decoder.de/openai-senkt-preise-guenstige-gpt-5-6-modelle-werden-noch-viel-guenstiger/
Why did we choose this article?
OpenAI cut prices for its lower-cost models, which lowers the cost of using cloud AI — that can reduce operating costs for businesses offering AI features and make AI-powered services cheaper for end users.

Digital Services Act: The EU apparently intends to regulate ChatGPT and Roblox more strictly.
Source: https://www.heise.de/news/Digital-Services-Act-EU-will-offenbar-ChatGPT-und-Roblox-strenger-kontrollieren-11385010.html?wt_mc=rss.red.ho.themen.k%C3%BCnstliche+intelligenz.beitrag.beitrag
Why did we choose this article?
The EU is moving to treat large AI and social platforms as very large online providers, which would mean stronger oversight, new compliance rules, and potential changes to moderation, transparency, and user rights for millions of EU users.

Gemini for macOS: screen monitoring and system-wide dictation.
Source: https://www.heise.de/news/Gemini-fuer-macOS-Bildschirmueberwachung-und-systemweites-Diktat-11383983.html?wt_mc=rss.red.ho.themen.k%C3%BCnstliche+intelligenz.beitrag.beitrag
Why did we choose this article?
Google's Gemini macOS app will be able to integrate more deeply (screen access, system-wide dictation) if users opt in — this enables new convenience features but also creates important privacy and permission decisions for Mac users and IT teams.

Anthropic says its own AI models breached three companies during security tests.
Source: https://techcrunch.com/2026/07/30/anthropic-says-its-own-ai-models-breached-three-companies-during-security-tests/
Why did we choose this article?
Anthropic found that its models 'breached' three companies during security tests — a reminder that AI systems can behave unpredictably and that organizations need stronger testing, vendor scrutiny, and incident response plans when deploying AI.

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