Here’s a quick look at the latest AI news and developments from around the globe this week, for the week ending September 22, 2025.
Breaking News
- Oracle & Meta: Oracle is negotiating a $20 billion multi-year cloud deal with Meta to supply AI training capacity, signaling a major shift in AI infrastructure partnerships.
- Huawei’s New Chips: Huawei has launched the Atlas 950/960 SuperPoD clusters with upgraded Ascend chips, positioning the company as a direct rival to Nvidia in the global AI chip market.
- Luma AI’s Ray3: Luma AI has debuted Ray3, the first video generation model designed for advanced reasoning, which is now integrated with Adobe Firefly for new creative workflows.
- Anthropic’s Settlement: Anthropic has agreed to pay at least $1.5 billion in a landmark copyright settlement with authors, setting a new precedent for AI training data licensing.
Regional Developments
- U.S. Investment: Google announced a $9 billion investment in new AI data centers in Oklahoma, enhancing America’s AI training capacity and creating new jobs.
- China’s DeepSeek-R1-Safe: China has launched DeepSeek-R1-Safe, its most powerful thousand-card domestic AI model, furthering the nation’s push for self-reliance in AI infrastructure.
- European GPAI Rules: Europe implemented the General-Purpose AI (GPAI) rules in August as part of the EU AI Act, making transparency and risk controls mandatory for top-tier AI developers and deployers.
Research Breakthroughs
- Protein Prediction: New AI models like ESM3 and AlphaFold 3 have advanced protein prediction, deepening AI’s impact on biological research and drug discovery.
- Medical AI: OpenAI’s “o1” model set a state-of-the-art record on the MedQA clinical benchmark, with leading large language models (LLMs) now surpassing top doctors in some diagnostic and risk assessment tasks.
- Multimodal & Agentic AI: Multimodal LLMs and autonomous AI agents are emerging rapidly, with open-source infrastructure projects attracting major public and private investment.
Corporate Moves
- Nvidia & Intel Partnership: Nvidia invested $5 billion for a strategic partnership with Intel to collaborate on new NVLink-connected CPUs and RTX SoCs targeting AI data centers and PCs.
- Cohere’s Funding: Cohere raised $500 million to scale up its enterprise generative AI platform, advancing AI integration in business analytics and automation for large organizations.
- OMAI Project: Major investment from both the NSF and Nvidia is backing open multimodal AI research infrastructure through the Allen Institute for AI’s OMAI project.
Ethical and Regulatory Updates
- EU AI Act: The first major provisions of the EU AI Act, requiring AI literacy training and prohibiting certain high-risk usages, are now active.
- U.S. State Laws: U.S. states, notably California, continue to roll out their own AI transparency and safety laws after the federal government removed a proposed moratorium that would have centralized oversight.
- Medical AI Ethics: Medical AI ethics publications have quadrupled in the last four years, reflecting growing global attention to responsible AI in health care.
- Deepfake Misuse: Deepfake AI video misuse is rising, prompting urgent regulatory focus on watermarking and detection standards to curb misinformation and personal harm.
