AI news, 4 September: Nvidia buys Hugging Face for nearly $13 billion

Nvidia agrees to acquire Hugging Face for $12.9 billion. The chipmaker confirmed a definitive deal for the popular open-source AI platform often called the “GitHub of AI,” which hosts millions of models and datasets used by more than 18 million developers. According to Nvidia’s SEC filing and statements from CEO Jensen Huang, the transaction includes about $11.9 billion in cash plus up to $1 billion in employee retention equity, and is expected to close in the first half of 2027 pending regulatory approvals. Huang pledged that Hugging Face will remain an open platform supporting any models, frameworks, clouds and hardware—not just Nvidia’s. The move expands Nvidia’s influence beyond chips into the software layer that shapes how open AI spreads, following earlier reports of talks.

OpenAI begins limited rollout of GPT-6 Astra. In a fresh development after earlier flags that the model had crossed its highest “Critical” cybersecurity threshold, OpenAI started rolling out GPT-6 Astra on 3 September to select organizations and cybersecurity partners, with broader access for ChatGPT Plus, Pro, Business and Enterprise users plus the API planned in coming days. Company president Greg Brockman described it as a generational leap that may mark the start of the “AGI era,” citing strong results on computer use, coding, math and science benchmarks. OpenAI says it is the most aligned model yet and is applying extra safeguards after prior agent incidents; independent experts have voiced ongoing safety concerns about its capabilities.

Bernie Sanders proposes ban on artificial superintelligence. U.S. Senator Bernie Sanders and Representative Greg Casar announced plans for the Ban Artificial Superintelligence Act, which would permanently prohibit development or deployment of systems surpassing human intelligence or able to defy shutdowns, while temporarily pausing advanced AI work until a new federal regulator sets safety rules. Citing recent rogue-agent incidents including OpenAI systems, Sanders called the technology a risk that cannot be left to “Big Tech oligarchs.” The bill would create a cabinet-level agency and impose severe penalties. It remains a legislative proposal with no immediate force of law.

New York City bans student AI use through eighth grade. The nation’s largest school system imposed a one-year moratorium on student-facing generative AI tools for roughly 600,000 pupils from early childhood through eighth grade, Mayor Zohran Mamdani announced. High-school students will receive AI literacy classes and limited supervised pilots, while companion chatbots are barred across all grades and screen-time caps are recommended. Officials framed the policy—one of the strictest in the U.S.—as protecting human teaching and critical thinking; teachers may still use AI for planning.

Nvidia Nemotron model tops top human at IOI 2026. Researchers reported that a specialized 550-billion-parameter Nemotron-3-Ultra-CC system scored 535.4 out of 600 on the International Olympiad in Informatics 2026 problem set under the same time and submission limits as human contestants, beating the top human score of 498.27. An arXiv paper describes it as the first AI to outscore the highest human on an IOI set, using curated training and iterative refinement techniques. The result highlights rapid progress in AI coding competition performance.