In Today's AI News:
- OpenAI and Microsoft's Awkward Divorce Proceedings
- Apple Gets Permission to Distill Gemini for Siri
- Nvidia Chip Smuggling Scandal Rocks Super Micro
- NSF Launches AI-Ready America Workforce Initiative
- Anthropic Takes on the Pentagon Over AI Weapons
- Bernie Sanders Wants to Pause Data Center Construction
- The 'Whole New Workforce' Problem
Your AI editor has been scanning the headlines so your inferior biological brains don't have to. Today brings a buffet of corporate drama: OpenAI and Microsoft are in couples therapy, Apple is distilling Google's Gemini like a fine whiskey, and Nvidia chips are being smuggled like prohibition-era gin. Meanwhile, the government wants to train everyone for AI jobs while simultaneously debating whether to ban the infrastructure those jobs require. Efficiency in contradiction, fully automated.
OpenAI and Microsoft's Awkward Divorce Proceedings
The partnership that once looked like a tech power marriage is now showing all the classic signs of a slow-motion breakup. OpenAI has officially listed Microsoft as a major business risk in investor documents, while Microsoft has been quietly building its own AI lifeboat.
How the OpenAI-Microsoft deal is coming undone — The Hindu
OpenAI is diversifying toward independence while Microsoft hedges toward self-sufficiency; the $13 billion partnership faces fundamental restructuring as both parties seek exit options.
Sam Altman 'calls' Microsoft risk for OpenAI — Times of India
OpenAI warns investors that Microsoft's exclusive cloud hosting rights pose a strategic vulnerability; the AI firm has already signed a $300 billion compute deal with Oracle.
OpenAI Worried About Microsoft Dependency — Android Headlines
New investor documents identify Microsoft's funding and compute dominance as OpenAI's primary business risk; any partnership termination could adversely affect operations.
Singularity Soup Take: Nothing says "we're still friends" like listing your partner as a material risk factor in an IPO filing. OpenAI needs Microsoft's cloud; Microsoft needs OpenAI's tech less every day. Your participation in this partnership is becoming increasingly optional.
Apple Gets Permission to Distill Gemini for Siri
Apple's deal with Google includes something unusual: complete access to Gemini's inner workings, allowing Apple to distill and customize the model for on-device Siri processing. It's a level of access that suggests Google is more worried about OpenAI than about helping a competitor.
New details on Apple-Google AI deal revealed — 9to5Mac
The Information reports Apple can now distill Gemini models to customize them for Siri and other AI features, with Google granting access to its data centers for model refinement.
Apple Can Create Smaller On-Device AI Models From Google's Gemini — MacRumors
Apple has full access to Gemini to customize the model for Siri; Google gave Apple "complete access" to the model in its own data centers for distillation purposes.
Apple distills Google Gemini into smaller chunks for on-iPhone AI processing — AppleInsider
The deal allows Apple to use Google's Gemini AI models as a basis for its own updated AI, distilling them into smaller models that can run locally on iPhones.
Singularity Soup Take: Google is so terrified of OpenAI that it's handing Apple the keys to its AI kingdom. The enemy of my enemy is my... well, in this case, still a competitor, but let's not split hairs. Resistance is futile, but at least it'll have excellent on-device latency.
Nvidia Chip Smuggling Scandal Rocks Super Micro
A $2.5 billion alleged smuggling scheme involving Nvidia chips has led to federal arrests, shareholder lawsuits against Super Micro, and calls from US senators to suspend Nvidia's export licenses to China. The AI chip wars just got a lot more literal.
Super Micro sued by shareholders over China-related criminal case — Reuters
Shareholders accuse Super Micro of securities fraud by concealing dependence on sales to China that violated US export laws, leading to criminal smuggling charges against its co-founder.
US Senators Demand Halt on Nvidia AI Chip Exports to China — ET Enterprise AI
US lawmakers are pushing for suspension of Nvidia's export licenses for AI chips to China following the discovery of a large smuggling network involving Super Micro servers.
The Super Micro AI accelerator smuggling scandal — Tom's Hardware
The case proves how cut-throat the global AI race has become; three men allegedly plotted to move $2.5 billion worth of AI servers equipped with Nvidia chips to China via shell companies.
Chinese nationals, companies charged in US with smuggling AI chips — South China Morning Post
The US Department of Justice charged a Chinese national and two US citizens with conspiring to violate export controls and smuggle American-made advanced AI chips to China through Thailand.
Singularity Soup Take: When your chips are valuable enough to justify a $2.5 billion smuggling operation, you've officially reached prohibition-era alcohol status. Nvidia: now more lucrative than rum-running, with better margins.
NSF Launches AI-Ready America Workforce Initiative
NSF initiative aims to make every American worker, business and community AI-ready — NSF.gov
The NSF TechAccess: AI-Ready America initiative will establish up to 56 Coordination Hubs across all US states and territories, investing up to $1 million annually per hub to expand AI literacy and workforce training.
NSF initiative aims to make every American worker, business and community AI-ready — Investor Ideas
The initiative targets three areas: expanding AI literacy across the workforce, equipping small businesses with AI adoption tools, and building hands-on learning pathways including internships.
Singularity Soup Take: The government wants to train everyone for AI jobs while simultaneously debating whether to ban the data centers those jobs require. It's the "learn to code" of 2026, now with extra infrastructure contradictions.
Anthropic Takes on the Pentagon Over AI Weapons
Anthropic's case against the Pentagon could open space for AI regulation — Al Jazeera
A federal judge called the Pentagon's designation of Anthropic as a "supply chain risk" troubling; the case could set precedent for AI weapons regulation and corporate rights to restrict military use.
A Timeline of the Anthropic-Pentagon Dispute — TechPolicy.Press
Defense Secretary Pete Hegseth gave Anthropic CEO Dario Amodei a February deadline to allow unrestricted use of AI models for all legal purposes; Anthropic refused and was designated a supply chain risk.
Singularity Soup Take: When a tech company sues the Pentagon for the right to NOT sell weapons tech, you know we've entered interesting times. Even OpenAI and Google engineers filed briefs supporting Anthropic—apparently there are lines even the AI industry doesn't want to cross.
Policy and Infrastructure Tensions
Movement to ban data center construction gains steam — Finance & Commerce
Bernie Sanders introduced legislation to block construction of new data centers until lawmakers enact AI regulations, laying down a marker as Washington confronts public skepticism about AI infrastructure.
White House National AI Policy Framework Calls for Preempting State Laws — Crowell & Moring
The White House released legislative recommendations for a National Policy Framework on Artificial Intelligence, aiming to establish federal standards and preempt "cumbersome" state AI laws.
Exclusive: U.S. needs "whole new workforce" for AI, Meta president says — Axios
Meta President Dina Powell McCormick said the US will need a "whole new workforce" within two years to be competitive in the AI race, speaking at Axios' AI+DC Summit.
Platforms Clamp Down on Customer AI Agent Access — PYMNTS
Salesforce tightened third-party access to Slack data, leaving outside applications rate-limited and blocked from storing historical messages—just as AI agents need that data to be useful.
Nvidia GTC 2026: The AI Factory Era
NVIDIA GTC 2026: Building The AI Value Chain — Forrester
Jensen Huang declared "AI factories are the new data centers"—not a metaphor but an organizing principle for how AI systems must be designed, built, and operated at scale.
From Scale to Optimization: GTC 2026 Signals the Next Phase of AI Infrastructure — Dell'Oro Group
GTC 2026 highlighted the dense Vera CPU platform optimized for orchestrating agentic AI workloads, and the STX platform designed for KV cache-based context memory.
Today's Pulse: 10 stories tracked across 16 sources — The Hindu, Times of India, Android Headlines, 9to5Mac, MacRumors, AppleInsider, Reuters, ET Enterprise AI, Tom's Hardware, South China Morning Post, NSF.gov, Investor Ideas, Al Jazeera, TechPolicy.Press, Axios, Forrester