9th & 10th December - AI News Daily - OpenAI, Google, Microsoft Unite to Launch Agentic AI Foundation

Send a text

🌍 INAI • The Open AI Hub

The Intelligence Atlas → the world’s most comprehensive, open hub of AI knowledge. 2 Million+ tools, models, agents, tutorials & daily news—free for all, updated every day.

https://github.com/inai-sandy/inAI-wiki

AI News Daily — Dec 10, 2025 Summary

Top Highlights: Major AI companies (OpenAI, Anthropic, Microsoft, Google) launched the Agentic AI Foundation with Anthropic's MCP donated to Linux Foundation for standardized agent interoperability. The Pentagon deployed GenAI.mil on Google Gemini for 3M military personnel. India proposed mandatory AI training royalties for copyrighted content. OpenAI committed $4.6B to Australian GPU infrastructure and workforce upskilling. FDA approved AIM-NASH, the first AI tool for liver biopsy analysis.

New Tools: CTGT launched adjustable LLM guardrails without retraining. AWS released goal-driven agent builder with production observability. Google Workspace Studio enabled no-code AI automations with 90% faster drafting. Amazon deployed Autonomous Agents for unsupervised long-running tasks. iFixit introduced free FixBot repair assistant. Marble and EgoEdit expanded 3D generation and egocentric editing tools.

LLM Updates: Mistral released Devstral 2 (123B) and 24B code models with 256K context. Zhipu shipped GLM-4.6V open multimodal model. Abu Dhabi's Jais 2 (70B) advanced Arabic language support. OpenAI accelerates GPT-5.2 for improved speed and reasoning. OpenAI "Confession" feature adds self-assessment for transparency. Zhipu AutoGLM enables on-device smartphone control.

Research: ARC-AGI contamination investigation revealed training-evaluation overlap. Stanford's 2025 AI Transparency Index showed declining openness among leading labs. UK AI Security Institute conducted red-vs-blue interpretability exercises. OfficeQA introduced grounded enterprise task evaluations. SAPO proposed stable RL for large models. GRAPE unified positional encodings.

Industry & Policy: EU opened antitrust probe into Google's AI use of publisher content. IBM acquiring Confluent for $11B. LangChain published voice-agent architecture comparison. Stanford lecture covered transformer computational motifs. Waymo detailed autonomous data operations. Qdrant demonstrated 100K+ product image semantic search.

Discussions: AGI definition debates continue. Deep agents show promise but brittleness in multi-step reasoning. Shift toward procedural skills over heavy agent stacks. Interest in symbolic-neural hybrids and concerns about photorealism crowding experimentation.

Support the show

🌍 INAI • The Open AI Hub

The Intelligence Atlas → the world’s most comprehensive, open hub of AI knowledge. 2 Million+ tools, models, agents, tutorials & daily news—free for all, updated every day.

https://github.com/inai-sandy/inAI-wiki