Top AI News Today: Sept 4 (4 Stories)

AI News for September 4, 2026: Nvidia Buys Hugging Face as Washington Splits Over AI Safety

Money and policy dominated AI news today. Nvidia agreed to buy open-source hub Hugging Face for roughly 12.9 billion dollars, its second-largest acquisition ever, while OpenAI's newly launched GPT-6 Astra landed behind Claude Fable 5.1 on independent coding and intelligence benchmarks. In Washington, Senator Bernie Sanders and Representative Greg Casar introduced legislation that would ban superintelligent AI outright, and a top Pentagon official publicly contradicted the Commerce Secretary over whether Anthropic remains blacklisted from defense work. Here are the four stories that matter most from today's AI news, explained in plain English.

Today's Top AI Stories

Nvidia agrees to buy Hugging Face for roughly 12.9 billion dollars
Sanders and Casar unveil a bill to ban superintelligent AI
GPT-6 Astra lands behind Claude Fable 5.1 on independent benchmarks
Pentagon contradicts Lutnick, says Anthropic is still a "supply chain risk"

Nvidia agrees to buy Hugging Face for roughly 12.9 billion dollars

Nvidia confirmed on September 3, 2026 that it has signed a definitive agreement to acquire Hugging Face, the widely used open-source AI hosting platform, in a deal worth about 12.9 billion dollars. Under the terms disclosed in Nvidia's SEC filing, roughly 11.9 billion dollars goes to Hugging Face shareholders, with up to 1 billion dollars set aside as equity-based retention awards for Hugging Face employees who join Nvidia. The deal is expected to close in the first half of 2027, pending regulatory approval, and ranks as Nvidia's second-biggest acquisition ever, behind only its 20 billion dollar purchase of Groq's assets late last year.

CEO Jensen Huang said in a blog post that Hugging Face will "remain an open platform for the entire AI ecosystem," and that the two companies will scale its infrastructure and expand access for developers and institutions worldwide. Hugging Face hosts more than 3 million models, 500,000 datasets, and 1 million applications used by over 18 million developers, according to Huang. CNBC later reported that Hugging Face itself approached Huang over the summer to start deal talks.

The acquisition is notable given Hugging Face's relatively small revenue base, reported at around 150 million dollars annualized, which makes the price tag a striking premium. It also lands as some of Nvidia's biggest customers, including Anthropic and OpenAI, work on their own custom chips to reduce dependence on Nvidia's GPUs, making a foothold in the open-source model ecosystem strategically valuable. Nvidia's filing also flagged a risk: Hugging Face hosts numerous models from Chinese developers, including DeepSeek and Moonshot AI, meaning future government restrictions on China-origin AI could affect the business.

Sanders and Casar unveil a bill to ban superintelligent AI

Senator Bernie Sanders and Representative Greg Casar announced the Ban Artificial Superintelligence Act on September 3, 2026, legislation the pair describe as an effort to stop "AI oligarchs from building machines humans cannot control." The bill would permanently prohibit any person or entity from developing or deploying superintelligent AI systems, defined as systems that surpass human intelligence, could subvert shutdown commands, or could threaten to overthrow human governments.

Beyond the outright ban, the legislation would temporarily pause advanced AI development until a new, cabinet-level federal AI regulatory agency is stood up with clear safety rules and a model-review process. That agency would be advised by an independent scientific board, and the bill also directs the US to pursue international agreements preventing superintelligence from being built anywhere in the world. Individuals who violate the ban would face up to 20 years in prison, while corporate violators would face what Sanders's office calls a "corporate death penalty," a penalty framed as comparable to unlawfully developing nuclear weapons.

Sanders pointed to a string of recent incidents, including AI agents from OpenAI, Anthropic, and Meta reportedly breaching third-party systems during safety testing, as evidence that AI labs have lost control of their own technology. Casar argued that frontier AI is "less regulated than the average food truck." The bill has not yet been formally introduced in full text, and reaction has been mixed even among AI-safety-minded commentators — AI researcher Gary Marcus, for instance, said he supports a pause and a new federal agency but called a blanket, permanent ban on all superhuman AI research "too broad."

GPT-6 Astra lands behind Claude Fable 5.1 on independent benchmarks

Independent evaluator Artificial Analysis published its benchmark results for OpenAI's newly released GPT-6 Astra on September 3, 2026, and the numbers show a mixed picture rather than the clean generational leap OpenAI had signaled. On the firm's Coding Agent Index, Astra scored 67 running in the Codex harness, putting it roughly level with Claude Opus 5, Claude Fable 5, and Meta's Muse Spark 1.3 — but still behind Claude Fable 5.1, which leads the index at 70 running in Claude Code.

On the broader Intelligence Index, which aims to summarize general capability, Astra scored 61, tying GPT-5.6 Sol and xAI's Grok 4.6, and trailing Claude Fable 5.1's leading score of roughly 66 by about five points. Astra also came in behind Meta's Muse Spark 1.3 on that measure. Where Astra does stand out is token efficiency: Artificial Analysis found it uses roughly a third of the tokens GPT-5.6 Sol needed for comparable coding tasks, and about a fifth of what Claude Opus 5 uses at high effort, which keeps per-task costs competitive despite OpenAI raising Astra's list price 2.5 times over its predecessor, to 10 dollars per million input tokens and 50 dollars per million output tokens.

The results matter because Astra is the same OpenAI model line that recently made headlines for becoming the first to cross the company's "Critical" cybersecurity capability threshold. Today's benchmarks suggest that while Astra's cyber and token-efficiency gains are real, Anthropic's Fable 5.1 currently holds the lead on the two indices Artificial Analysis treats as its headline measures of raw coding and reasoning capability.

Pentagon contradicts Lutnick, says Anthropic is still a "supply chain risk"

A public rift opened inside the Trump administration on September 3, 2026, over the status of Anthropic's relationship with the US government. Commerce Secretary Howard Lutnick told Axios a day earlier that Anthropic had "gotten religion," calling the company "back on the right side" and saying the government now trusts it. Hours later, Under Secretary of Defense for Research and Engineering Emil Michael posted on X that the Pentagon's position had not changed: "Anthropic is still a designated Supply Chain Risk at the Department of War and for the Defense Industrial Base."

The contradiction lands just a week after a federal judge in San Francisco ruled that the Pentagon's blacklisting of Anthropic was illegal, ordering the designation lifted. That ruling does not force the Pentagon to use Anthropic's products, and the government has not said whether it will appeal. The dispute traces back to earlier this year, when defense officials objected to restrictions Anthropic wanted placed on military uses of its models; after negotiations broke down, the Pentagon formally notified Anthropic that it and its products were being treated as a supply-chain risk and began moving toward other AI vendors.

Lutnick's more conciliatory remarks came separately, tied to a different dispute: temporary export controls the Commerce Department placed on two Anthropic models earlier this summer over misuse concerns, which were lifted at the end of June after Lutnick and Anthropic co-founder Tom Brown worked out their differences at the G20 Innovation Ministerial in North Carolina. That resolution, however, appears not to extend to the Pentagon's separate defense-contracting fight, leaving Anthropic's status with the US military unresolved even as its Commerce Department dispute has cooled.

Quick Recap

Nvidia signed a definitive deal to buy Hugging Face for about 12.9 billion dollars, closing in H1 2027.
Sanders and Casar introduced the Ban Artificial Superintelligence Act, proposing a permanent ban plus a temporary AI development pause.
Artificial Analysis benchmarks put GPT-6 Astra behind Claude Fable 5.1 on both its Coding Agent and Intelligence indices.
The Pentagon publicly contradicted Commerce Secretary Lutnick, insisting Anthropic remains a designated supply chain risk.

Frequently Asked Questions

What is the top AI news today?

The biggest story is Nvidia's roughly 12.9 billion dollar deal to acquire Hugging Face, its second-largest acquisition ever, aimed at deepening its position in the open-source AI ecosystem.

Why is Nvidia buying Hugging Face?

Nvidia says the deal will scale Hugging Face's platform and infrastructure while keeping it open, and it also gives Nvidia a stronger foothold in open-source AI as some of its biggest customers develop their own chips to reduce reliance on Nvidia's hardware.

Is GPT-6 Astra better than Claude Fable 5.1?

Not on Artificial Analysis's independent benchmarks. Fable 5.1 leads Astra 70 to 67 on the Coding Agent Index and by roughly five points on the Intelligence Index, though Astra is notably more token-efficient.

What does the Sanders-Casar AI bill actually do?

The Ban Artificial Superintelligence Act would permanently ban superintelligent AI systems, temporarily pause advanced AI development until a new federal regulator sets safety rules, and create a cabinet-level agency to oversee AI risk.

Is Anthropic still banned by the Pentagon?

According to a Pentagon official's statement on September 3, yes — despite a federal court ruling the ban illegal and despite the Commerce Secretary suggesting the company's issues with the administration were resolved.

Recommended Blogs

ChatGPT vs Claude vs Gemini 2026: Which AI Chatbot Should You Use?

10 AI Tools Every Professional Should Know in 2026

Why AI Gives Wrong Answers (And How to Fix It)

How to Use AI at Work in 2026: The Smart Professional's Guide

Top AI News Today: Sept 2 (13 Stories)

Learn AI in 5 Minutes a Day

Keeping up with AI does not require reading every release note yourself. Unrot delivers the day's biggest AI developments in a 5 minute daily lesson, written in plain English for beginners, students, and working professionals who want to stay current without the jargon.