Trump Cancels AI Executive Order, Bans Anthropic From Government — OpenAI Rushes In
In a dramatic evening of AI policy, President Trump has simultaneously postponed signing a major AI executive order and ordered all federal agencies to stop using Anthropic technology. The executive order — which would have established federal AI safety standards and testing requirements — was shelved after intense pressure from Silicon Valley and concerns it could weaken the U.S. tech edge. Separately, Trump ordered all military contractors and federal agencies to cease business with Anthropic, citing a standoff over AI safety and the Pentagon’s refusal to accept Anthropic’s usage restrictions. Hours later, OpenAI struck a deal with the Pentagon, announcing it would provide classified AI capabilities to the Department of Defense. OpenAI CEO Sam Altman defended the move, while a top researcher resigned in protest. The EFF called it a “weasel words” deal that won’t stop AI-powered surveillance. (Sources: NYT, AP News, BBC, Politico, CNBC, CNN, NPR)
Nvidia Launches Cosmos 3: Open Foundation Model for Physical AI
At Computex Taipei, Nvidia unveiled Cosmos 3, an open frontier foundation model for physical AI built on a Mixture-of-Transformers architecture. The 32B-parameter model is designed to teach robots, autonomous vehicles, and industrial systems to reason about the physical world — combining vision, audio, and action planning into a single open model. Cosmos 3 is available on GitHub under a permissive license, alongside reference implementations for world model training, action planning, and physical AI reasoning pipelines. Nvidia also debuted its RTX Spark “superchip” for Windows laptops (covered in our earlier edition), positioning the company across both the training and inference layers of the AI stack. (Sources: NVIDIA Newsroom, HPCwire, Axios, Stock Titan)
Florida Sues OpenAI and Sam Altman, Seeks Personal Liability
The State of Florida has filed a sweeping lawsuit against OpenAI and its CEO Sam Altman, alleging the company knowingly marketed ChatGPT despite “utter disregard for the risk to human life.” The Florida Attorney General’s suit claims OpenAI concealed serious safety risks — including the potential for ChatGPT to fuel violence, generate harmful content for children, and bypass safeguards. Notably, the suit seeks to hold Altman personally liable, a rare move that could set a precedent for executive accountability in AI. Florida joins a growing wave of state-level legal action against AI companies. (Sources: CNBC, NPR, CNN, NYT, Variety)
Bernie Sanders Proposes 50% Public Ownership of AI Companies
Senator Bernie Sanders has proposed sweeping legislation that would grant the American public a 50% ownership stake in the largest U.S. AI companies. In a New York Times op-ed, Sanders argued that “A.I. is a public resource — you should own half of it.” The bill would create an American AI Sovereign Wealth Fund funded by a 50% equity tax on frontier AI labs, with proceeds directed toward public education, healthcare, and infrastructure. The proposal has drawn sharp reactions from Silicon Valley, with critics calling it a “seizure” of private property while supporters point to AI’s reliance on publicly funded research and data. (Sources: NYT, Mashable, Washington Examiner, Crypto Briefing)
Meta Delays ‘Avocado’ AI Model After Performance Concerns
Meta has delayed the rollout of its next-generation AI model, codenamed “Avocado,” following disappointing internal benchmark results. According to NYT and Reuters, the model’s performance fell short of GPT-5.5 and Gemini 3.5, prompting Meta to consider an unprecedented fallback: licensing Google’s Gemini for its products. The delay puts Meta’s $135 billion AI bet under scrutiny, as the company has already spent an estimated $14 billion building an AI super-team. The move marks a significant shift — Meta has historically relied entirely on open-source and in-house models (Llama series) rather than licensing from rivals. (Sources: NYT, Reuters, Fortune, CNET, Investor’s Business Daily)
White House Considers Pre-Release Vetting of AI Models
The Trump administration is moving toward requiring government testing and approval of advanced AI models before they can be released to the public, according to multiple reports. The proposal — reported by NYT, Bloomberg, Politico, and Axios — would apply to frontier models from companies including Google, Microsoft, OpenAI, Anthropic, and xAI. A coalition of 60+ MAGA allies has publicly urged Trump to support pre-release AI vetting. The proposal marks an unexpected turn for an administration that had campaigned on deregulation, reflecting growing bipartisan concern over AI safety. (Sources: NYT, Bloomberg, Politico, Axios, CNBC)
In Brief
- Alphabet plans to raise $80 billion through stock sales to fund its AI infrastructure buildout. Berkshire Hathaway will purchase $10 billion of the newly issued shares, signaling Warren Buffett’s confidence in Google’s AI strategy. (Sources: CNBC, Reuters, WSJ)
- OpenAI has opened GPT-5.5 Cyber to UK banks after Anthropic blocked access to its Mythos model for cybersecurity purposes. Japan’s banks have also received access. The move escalates the AI cold war between the two frontier labs. (Sources: BBC, Cybernews, Reuters)
- Moonshot AI launched Kimi K2.6, its latest 1T-parameter open-source model, challenging Anthropic’s Claude series. The company also raised $2 billion at a $20 billion valuation, underscoring voracious demand for open-source alternatives. (Source: TechCrunch)
- Tencent pledged higher AI investment in 2026 after U.S. chip curbs impacted its capex plans, while BMW partnered with Mistral AI to use AI in crash simulation — replacing costly physical crash tests with AI-driven simulations. (Sources: Reuters, Engineer Live)
- DuckDuckGo reported a 30% surge in installations as users push back against Google’s AI-driven search redesign. Its “no-AI” search engine is now seeing record traffic. (Source: TechCrunch)
⏰ Limited-Time Free Models
Several models remain free on OpenRouter for a limited time:
| Model | Free Until | Context | Standard Pricing |
|---|---|---|---|
| Qwen3 30B A3B (Qwen) | June 5 (2 days) | 131K | $0.09/M tokens prompt, $0.45/M tokens completion |
| Llama 3 Euryale 70B v2.1 (Sao10k) | June 5 (2 days) | 8K | $1.48/M tokens |
| Hermes 2 Pro - Llama-3 8B (NousResearch) | June 5 (2 days) | 8K | $0.14/M tokens |
| Claude Opus 4.6 (Fast) (Anthropic) | June 29 (26 days) | 1M | $30/M tokens prompt, $150/M tokens completion |
🤗 Notable HuggingFace Uploads
The HuggingFace ecosystem showed quiet activity over the past hours, with most uploads remaining experimental.
- NN-Dataset/tflite — TensorFlow Lite dataset collection with 16,839 downloads.
- Miiche/visualrl-ft-r1p9 — Qwen2.5-VL fine-tune for visual reinforcement learning (264 downloads).
- geonho1/Mistral-7B-Instruct-v0.2-4b-r64-task319 — LoRA adapter for Mistral 7B, part of the “lots-of-loras” collection.
🚀 New on OpenRouter
- MiniMax M3 (MiniMax) — 1M context, priced at $0.30/M tokens prompt, $0.30/M tokens completion. Features sparse attention for 15.6× faster long-context decoding.
⭐ GitHub Trending: AI Edition
| Repo | Stars | Description |
|---|---|---|
| pewdiepie-archdaemon/odysseus | ⭐20,014 | Self-hosted AI workspace with MCP server support |
| GordenSun/GordenPPTSkill | ⭐1,384 | AI-friendly PPT builder with 17 Chinese PPTX templates |
| withkynam/vibecode-pro-max-kit | ⭐699 | Spec-driven coding harness for AI-assisted development |
| 2aronS/Duel-Agents | ⭐637 | CLI, SDK, and IDE plugins for multi-agent systems |
| asz798838958/aBaiAutoplus | ⭐613 | Multi-platform AI account auto-registration and management |
💡 Key Trends
This evening’s news reveals three fault lines reshaping AI in real-time. Government vs. frontier labs: Trump’s simultaneous AI EO cancellation, Anthropic ban, and OpenAI Pentagon deal show how quickly the federal AI landscape can shift — with companies gaining or losing government access in hours. The $300B+ infrastructure wave: Alphabet ($80B), SoftBank (€75B), and Tencent joined the capex race, with Berkshire Hathaway making its first major AI infrastructure bet through Alphabet. Physical AI goes open: Nvidia’s Cosmos 3 release signals that the next frontier — AI that can reason about and act in the physical world — is being open-sourced at the foundation model level, potentially accelerating robotics and autonomous systems the way Llama accelerated language AI.