In this news and research episode, Ray and Dan unpack a whirlwind of global developments in AI and education. From major US announcements, like Microsoft, Google, and Amazon offering free AI tools and training for students, to Australia's push for sovereign AI infrastructure, it’s clear the AI education landscape is shifting fast. They explore the massive copyright settlement involving Anthropic and the controversial Books3 dataset, dig into what AI is actually trained on, and consider the implications of training data transparency. They also spotlight Georgia Tech’s Jill Watson project and a new study comparing different RAG (Retrieval-Augmented Generation) strategies (essential reading for anyone building AI tutors or educational bots?) Plus: Google’s AI Quests, Australia’s new social media ban for under-16s, OpenAI’s new certifications, and a growing global interest in culturally specific AI models.   News - Links   White House AI Education Task Force https://www.theverge.com/policy/772084/amazon-google-microsoft-white-house-ai-education Key tech company commitments announced:

• Microsoft: https://blogs.microsoft.com/on-the-issues/2025/09/04/new-white-house-commitments/ 
• Google:  https://ai.google/literacy  https://blog.google/outreach-initiatives/education/ai-education-efforts/ 
• Amazon: https://www.aboutamazon.com/news/policy-news-views/how-amazon-is-supporting-the-white-houses-ai-education-initiatives 

  Digital Education Council's "AI in the Workplace 2025" report https://www.digitaleducationcouncil.com/post/ai-in-the-workplace-2025    Microsoft becomes best buddies with Anthropic https://www.computerworld.com/article/4054738/microsoft-to-tap-anthropic-for-office-365-as-enterprises-weigh-risks-of-ai-lock-in.html  https://developers.slashdot.org/story/25/09/17/1927233/microsoft-favors-anthropic-over-openai-for-visual-studio-code    Anthropic settled copyright lawsuit https://www-pbs-org.cdn.ampproject.org/c/s/www.pbs.org/newshour/amp/nation/anthropic-to-pay-authors-1-5b-in-landmark-settlement-over-pirated-chatbot-training-material    OpenAI announces new certification  https://openai.com/index/expanding-economic-opportunity-with-ai/    Greece give ChatGPT Edu to high school students https://www.yahoo.com/news/articles/greece-openai-agree-deal-boost-150237148.html    OpenAI's Global Faculty AI Project https://academy.openai.com/public/resources/the-global-faculty-ai-project-2025-09-04  Some of the ideas that are there include:

• Brinnae Bent, PhD (Duke University) on “Hack Your Grade,” an experiential assignment where students try to outsmart a chatbot, building AI literacy and critical thinking through hands-on practice.
• David J. Malan (Harvard University) on creating a “virtual rubber duck” (Dan, do you know what 'rubber ducking' is?) debugging system that guides computer science students instead of giving them the answers.
• andre j. hermann (Houston City College) on using AI in the photo studio to design real-world assignments that teach both technical craft and career skills like creative briefs, brainstorming, and execution.
• Marcos Rojas Pino, MD (Stanford University) on Clinical Mind AI, which grew from a custom GPT into a multilingual simulation platform to run realistic patient encounters, making high-quality clinical reasoning training accessible to all health professions students.

Google launches AI Quests to teach AI literacy to students aged 11 to 14 https://blog.google/outreach-initiatives/education/ai-quests/  https://research.google/ai-quests/intl/en_gb    Australian eSafety changes on social media (and potentially AI) https://www.afr.com/politics/federal/age-assurance-technology-needed-to-protect-teens-from-ai-chatbots-20250909-p5mtln    Sovereign Australia AI launch https://www.afr.com/technology/we-can-do-it-for-under-100m-start-up-joins-race-to-build-local-chatgpt-20250908-p5mt5o    Research - Links   Georgia Tech’s Jill Watson Outperforms ChatGPT in Real Classrooms https://research.gatech.edu/georgia-techs-jill-watson-outperforms-chatgpt-real-classrooms    Aligning LLMs for the Classroom with Knowledge-Based Retrieval -- A Comparative RAG Study https://arxiv.org/abs/2509.07846