Major Updates
AI today delivered a mix of legal victories, corporate breakups, and infrastructure shifts. A federal judge voided the Pentagon's blacklist of Anthropic, OpenAI is severing ties with Cursor after SpaceX's $60B acquisition, and Nvidia paused its massive $36B cloud financing program amid antitrust concerns. Sony and Warner Chappell sued Anthropic over training data, while Tencent open-sourced a 770B-parameter model and Anthropic quietly raised Claude Code prices by 17%.
Judge Strikes Down Pentagon's Anthropic Blacklist as Illegal
U.S. District Judge Rita Lin ruled Thursday that the Pentagon's designation of Anthropic as a supply-chain risk was unlawful, calling it First Amendment retaliation against CEO Dario Amodei's refusal to allow unrestricted military use of Claude, including mass surveillance and fully autonomous weapons. The 59-page order states officials sought to "make a public example out of Anthropic." The White House said it will appeal the decision.
The ruling ends months of friction over what the U.S. military can do with Claude. Anthropic had argued the blacklist violated its free-speech rights after Amodei publicly opposed certain defense applications. This is a landmark case for AI companies' free-speech protections when their executives publicly disagree with government clients over model deployment.
Sources: Axios, NPR, Fortune, The Next Web
OpenAI Cuts Off Cursor After SpaceX $60B Buyout
OpenAI notified SpaceX on August 28 that it will wind down its contract supplying models to Cursor on November 12, invoking a change-of-control clause triggered by SpaceX's $60 billion acquisition of Anysphere. OpenAI cited past contract violations by Musk-owned companies (Twitter and xAI) and internal evaluations showing its Astra model's agentic coding capabilities could constitute critical cyber capabilities. Cursor will retain access to Anthropic, Google, and SpaceXAI models after the cutoff.
The decision means developers using GPT-powered features in Cursor have until mid-November to adapt. Cursor co-founder Michael Truell confirmed OpenAI represents only about 5% of Cursor's traffic. The move highlights how Elon Musk's corporate empire continues to reshape the AI tooling landscape, forcing developers to reconsider their stack dependencies.
Sources: CNBC, OpenAI blog, Cybersecurity News, Fire The Ring
Nvidia Pauses $36B AI Cloud Financing Over Antitrust
Nvidia has paused its AI Compute Partnership less than two months after launch, the Wall Street Journal reported. The program guaranteed GPU rentals to smaller cloud providers in exchange for 50% of revenue above a base hourly rate and had accumulated $36B in commitments per Nvidia's quarterly filing. Employees warned that the arrangement could invite antitrust scrutiny given how much control Nvidia gained over its own customers' businesses.
The pause raises questions about whether Nvidia can simultaneously be the dominant chipmaker and the financier of its customers' infrastructure. Partners balked at customer-approval terms that gave Nvidia significant leverage over how rented GPUs were deployed. This comes as regulators globally scrutinize Nvidia's growing influence over the AI supply chain.
Sources: WSJ, Business Model Analyst
Sony, Warner Chappell Sue Anthropic Over Claude Training Data
Sony Music Publishing and Warner Chappell sued Anthropic, CEO Dario Amodei, and co-founder Benjamin Mann in Northern District of California on August 28, alleging that "tens of thousands" of copyrighted compositions were torrented from Library Genesis and Pirate Library Mirror and scraped from MusixMatch and LyricFind to train Claude. The complaint seeks up to $150,000 per work for willful infringement plus $25,000 per removal of copyright management information, naming songs like "Ain't No Mountain High Enough," "All I Want for Christmas is You," and "Eye of the Tiger."
The lawsuit represents one of the largest potential copyright exposures in AI training, with potentially multi-billion-dollar liability. It follows a pattern of rising copyright litigation against AI companies and raises questions about whether the fair-use doctrine will hold as generative AI models trained on vast copyrighted datasets begin generating near-identical outputs.
Sources: The Verge, The IT Guys Fix
Tencent Open-Sources 770B Hy4 Model with 1M-Token Context
Tencent's Hunyuan team released Hy4-preview on Hugging Face under Apache 2.0, featuring 770 billion total parameters with 49 billion activated per token, 256 routed experts plus 1 shared, and a 1-million-token context window. The model card claims a 92.3 score on GPQA Diamond and 65.7 on SWE-bench Pro, with FP8 weights and vLLM/SGLang Docker recipes shipped day one. API pricing starts at $0.834 per million input tokens and $2.501 per million output tokens on Tencent Cloud TokenHub.
This release adds a major Chinese contender to the open-weights frontier, offering competitive benchmark performance at a fraction of closed-source API costs. The 1M-token context window puts it in the same league as Gemini and Claude for long-document processing, and the open-source licensing could accelerate adoption in research and enterprise settings.
Sources: AI Weekly, Hugging Face
Anthropic Claude Code Hike Is Actually a 17% Net Cut
Anthropic told Claude Code users on August 29 that it will "permanently" raise standard weekly limits by 25% for Pro, Max, Team, and seat-based Enterprise plans starting September 14 — but the current temporary 50% boost ends at that point, making the net effect a 17% reduction from today's usable cap. Developer backlash was swift, with users noting the messaging was deliberately misleading: "you read the first half of the message, you think it's good news, you read the second half..."
The move follows a pattern of Anthropic adjusting pricing and rate limits as it balances infrastructure costs with user growth. For heavy Claude Code users, the effective reduction means either paying more for higher tiers or accepting slower workflows, potentially driving some teams to evaluate alternatives like Cursor, Copilot, or Codex.
Sources: AI Weekly, Anthropic blog
Claude Sonnet 5 Price Rises to $3/M on August 31
Anthropic's Claude Sonnet 5 will increase from $2 to $3 per million input tokens and from $10 to $15 per million output tokens on August 31, ending a promotional pricing period that lasted roughly two months. The standard rates position Sonnet 5 still well below Opus 4.8, which costs $5 per million input and $25 per million output, but the rollback signals that early-adopter pricing was never intended to be permanent.
The price increase comes as Anthropic scales infrastructure to meet growing demand. For API-heavy applications, the $1/M increase on input tokens translates to meaningful cost differences at scale — a service processing 10 billion tokens monthly would see its bill rise by roughly $10,000 per month.
Sources: BenchLM.ai, Anthropic pricing page, The Automated
Trump Weighs Chip Tariffs on Laptops, Servers, and Consoles
The Trump administration is preparing broader semiconductor tariffs that would extend beyond raw chips to finished goods built with foreign silicon — including laptops, gaming consoles, and data-center servers, according to CNBC and Politico. Commerce Secretary Howard Lutnick's framework would tie duty-free import allowances to how much domestic fab capacity a company commits to building, with January's 25% AI-chip tariff serving as the baseline.
The proposed tariffs could reshape the entire consumer electronics supply chain, raising prices on everything from gaming consoles to enterprise servers. The policy aims to accelerate domestic semiconductor manufacturing but risks increasing costs for consumers and businesses that depend on imported hardware. Analysts warn the tariffs could also complicate Nvidia's and AMD's relationships with global manufacturing partners.
Sources: CNBC, Politico, AI Weekly
General Intuition Nearly Triples to $6B on World Models
World-model startup General Intuition is raising at a $6 billion pre-money valuation from new investors Valor Equity Partners, Point72 Ventures, and Seven Seven Six, nearly tripling the $2.3 billion mark set just eight weeks ago in a $320 million Series A. The New York startup, spun out of gameplay-clip platform Medal in October 2025, trains world models on hundreds of millions of hours of video-game footage and has adapted its model to physical navigation tasks with just 8 minutes of real-world data.
The rapid valuation escalation reflects surging investor interest in world models as a foundational AI paradigm. Existing backers Khosla Ventures and General Catalyst are re-upping, and the new capital is earmarked for pushing the model into robotic embodiments via CoreWeave compute. The company's founders previously walked away from a Prometheus offer of a $1-2M salary, 35% stake, and $2B in committed financing — a bet that looks prescient as world models emerge as a distinct category.
Sources: TechCrunch, PYMNTS, Crypto Briefing
Nvidia Groq 3 LPX Inference Rack Ships, Nebius Is First Customer
Nvidia said its Groq 3 LPX, the dedicated inference accelerator built from its $20B Groq acqui-hire, has entered full production and slots into the Vera Rubin platform with up to 256 LPX accelerators per rack. Nebius will be the first cloud customer, deploying LPX racks alongside Vera CPUs and Rubin GPUs in its Token Factory. A benchmark from Artificial Analysis clocks 3,400 output tokens per second on Gemma 4 31B at 100K context. SpaceX said its next-generation AI stack, including orbital data centers, will run on Nvidia's Vera CPUs.
The Groq 3 LPX represents Nvidia's most aggressive move yet into the inference market, where custom silicon from Broadcom and Cerebras has been gaining ground. With 256 accelerators per rack and compatibility with the Vera Rubin platform, Nvidia is betting that integration with its existing CPU and GPU ecosystem will win over customers who might otherwise choose point-solution inference hardware.
Sources: AI Weekly, Nvidia blog, Artificial Analysis
AM Intelligence Orders 9,000 Vera Rubin Systems for $8B Buildout
Greenko Group's Hyderabad-based AM Intelligence has placed a binding order for approximately 9,000 Nvidia Vera Rubin NVL72 rack-scale systems for delivery in Q1 2027, positioning it as one of Asia's first frontier Vera Rubin clusters. The company plans roughly $8 billion in capex to bring 200MW online near-term, scaling toward 1GW of compute-as-a-service across India, the US, Finland, and Malaysia. Founder Mahesh Kolli says a US customer has already reserved initial capacity, and the Hyderabad facility is engineered to deliver about 450 exaFLOPS of NVFP4 inference compute backed by Greenko's low-cost renewable power.
This order underscores the massive scale of AI infrastructure buildout in Asia, where governments and private capital are racing to build gigawatt-scale data centers. The deal also highlights Nvidia's Vera Rubin platform as the next-generation standard for frontier AI training and inference, with major customers locking in supply years before production ramps.
Sources: AI Weekly, TechCrunch
California Carves Linux, BSDs Out of Age-Verification Law
California's Assembly passed AB-1856 in a 69-0 vote, exempting software distributed under GPL, MIT, BSD, or Apache licenses from the Digital Age Assurance Act — sparing Debian, Fedora, Ubuntu, Arch, SteamOS, and the BSD family from being forced to collect user ages at OS setup. A second provision exempts non-standalone package-manager components. Governor Newsom, who signed the original act last October, now decides whether to sign the exemption into law.
The unanimous vote reflects growing bipartisan recognition that age-verification mandates could undermine open-source infrastructure that underpins the internet. The bill specifically protects the package managers and libraries that developers rely on daily, addressing concerns that mandatory age gates at the OS level could break automated workflows and expose user data to new attack surfaces.
Sources: AI Weekly
Frequently Asked Questions
Q: What did the judge rule about the Pentagon's blacklist of Anthropic?
A: U.S. District Judge Rita Lin ruled the Pentagon's designation of Anthropic as a supply-chain risk was unlawful and constituted First Amendment retaliation against CEO Dario Amodei's public criticism of unrestricted military AI use. The White House plans to appeal.
Q: Why is OpenAI cutting off Cursor?
A: OpenAI invoked a change-of-control clause after SpaceX acquired Anysphere (the company behind Cursor) for $60 billion. OpenAI cited past contract violations by Musk-owned companies and concerns that its Astra model's agentic capabilities could pose critical cyber risks. Cursor will keep Anthropic, Google, and SpaceXAI models.
Q: Why did Nvidia pause its $36B AI cloud financing program?
A: Nvidia paused the AI Compute Partnership after partners raised concerns about customer-approval terms that gave Nvidia excessive control over rented GPU infrastructure. The program had accumulated $36B in commitments in under two months, and employees warned it could attract antitrust scrutiny.
Q: What is the Sony and Warner Chappell lawsuit against Anthropic about?
A: Sony and Warner Chappell allege Anthropic trained Claude on tens of thousands of copyrighted songs scraped from sites like Library Genesis and MusixMatch. The lawsuit seeks up to $150,000 per work for willful infringement, potentially exposing Anthropic to multi-billion-dollar liability.
Q: How much does Claude Sonnet 5 cost after August 31?
A: After August 31, Claude Sonnet 5 pricing increases from $2 to $3 per million input tokens and from $10 to $15 per million output tokens. This ends a promotional rate that lasted roughly two months since the model's launch on June 30.
Q: What is the California AB-1856 law about?
A: AB-1856 exempts open-source software licensed under GPL, MIT, BSD, or Apache from California's Digital Age Assurance Act, which otherwise requires age verification at OS setup. The unanimous 69-0 Assembly vote protects Linux distributions, package managers, and the broader open-source ecosystem.
Q: Why is Anthropic Claude Code's new pricing actually a 17% cut?
A: Anthropic announced a 25% permanent increase in weekly limits starting September 14, but simultaneously ended a temporary 50% boost, resulting in a net 17% reduction in usable capacity from current levels. Developers criticized the messaging as deliberately misleading.
Sources
All sources cited throughout this roundup were retrieved from August 28-30, 2026, via web search, official company announcements, and the AI Weekly aggregator (aiweekly.co/ai-news-today). For detailed reporting on each story, see the source links embedded above.
Top comments (0)