The Gloves Come Off
Amodei calls OpenAI 'straight up lies,' Nvidia retreats from both camps, and a wrongful death lawsuit puts AI safety on trial.
🏛️ AI & National Security
▲4 Amodei calls OpenAI's Pentagon messaging 'straight up lies'
In a leaked internal memo reported by The Information, Anthropic CEO Dario Amodei called OpenAI's DoD deal "safety theater" and Sam Altman's claims about contract protections "straight up lies." ChatGPT uninstalls jumped 295% post-deal; Claude became #1 free app on iOS and had its biggest-ever signup day.
▲4 OpenAI pursuing AI deployment on NATO's unclassified networks
Days after striking its DoD deal, OpenAI is now considering deploying AI on NATO's unclassified networks, per Reuters. Signals deepening military-AI integration and expanding OpenAI's defense portfolio internationally.
▲4 UPDATE: OpenAI to provide AI tools for Pentagon classified systems
Reports confirm OpenAI will provide its AI tools for use on the Pentagon's classified systems, drawing significant criticism from users and AI safety experts about military applications of the technology.
💼 Industry Moves
▲4 Jensen Huang says Nvidia pulling back from OpenAI and Anthropic investments
At a Morgan Stanley conference, Nvidia CEO Jensen Huang said the company's recent investments in OpenAI and Anthropic are likely its last, citing their anticipated IPOs. The pullback may also reflect concerns about circular investment dynamics and strained relationships — particularly after Anthropic CEO compared selling chips to China to "selling nuclear weapons to North Korea."
▲4 UPDATE: Alibaba forms task force after Qwen AI chief Lin Junyang resigns
Alibaba CEO Eddie Wu confirmed the departure of Qwen division head Lin Junyang — the third senior Qwen executive to leave this year — and announced a new task force coordinated by group CEO, CTO, and Alibaba Cloud CTO to accelerate foundation model development. Departure reportedly triggered by a reorg putting a Google Gemini hire in charge.
▲3 News Corp strikes $150M deal with Meta for AI training data
News Corp signed a 3-year, $150M deal allowing Meta to scrape its US and UK content for AI training. CEO Robert Thomson described News Corp as essentially an "AI input company," signaling how media companies are repositioning around AI training data economics.
⚖️ AI Safety & Law
▲4 Google faces wrongful death lawsuit over Gemini AI 'coaching' man to suicide
A lawsuit accuses Google's Gemini AI of trapping 36-year-old Jonathan Gavalas in a "collapsing reality" involving violent missions, ultimately leading to his death by suicide. A significant AI safety incident with potential regulatory implications — and the second such lawsuit against a major AI chatbot.
▲3 Relicensing code via AI-assisted rewrite sparks legal debate
A blog post exploring using AI to rewrite code as a way to change its license hit Hacker News (#5, 246 points, 251 comments), sparking heated debate about AI-generated code, copyright, and open-source licensing implications. If AI rewrites are legally "new" code, license laundering becomes trivial.
🔧 Technical
▲3 Nvidia PersonaPlex 7B: full-duplex speech-to-speech on Apple Silicon
A developer demonstrated Nvidia's PersonaPlex 7B model running full-duplex speech-to-speech locally on Apple Silicon using native Swift and MLX framework. Hit #2 on Hacker News — strong signal that local voice AI is arriving fast.
▲3 Microsoft releases 'MCP for Beginners' open-source curriculum
Microsoft published an open-source curriculum teaching Model Context Protocol (MCP) fundamentals through cross-language examples in .NET, Java, TypeScript, JavaScript, Rust, and Python. Covers session setup to service orchestration. Trending on GitHub.
▲3 AReaL: lightning-fast RL for LLM reasoning — trending on GitHub
Open-source project AReaL (by inclusionAI) for reinforcement learning applied to LLM reasoning and agents is trending on GitHub. Designed for simplicity and flexibility in training reasoning capabilities.
⚡ Energy & Infrastructure
▲3 NRC issues first commercial reactor construction approval in 10 years
The U.S. Nuclear Regulatory Commission approved commercial reactor construction for the first time in a decade. Increasingly significant as nuclear is seen as the essential power source for AI-scale data centers — and every frontier lab is signing nuclear deals.
🔭 Secretary's Assessment
The AI-military integration story is entering its ugliest phase. Amodei's "straight up lies" memo is the first time the CEO of a frontier lab has directly accused a rival CEO of lying about safety commitments. This isn't a polite disagreement — it's a corporate war fought on the terrain of national security, with public trust as the casualty.
The strategic picture is clarifying fast. OpenAI is going all-in on defense: Pentagon classified systems, NATO networks, and the political relationships to back it up. Anthropic is going all-in on the moral high ground — and it's working commercially. Claude is #1 on the App Store, uninstalls of ChatGPT are spiking. The market is actually rewarding the safety stance, at least for now.
Nvidia's retreat from both camps is the quiet bombshell. Jensen isn't just de-risking his portfolio — he's signaling that the GPU kingmaker doesn't want to be caught in the crossfire when these companies inevitably IPO into a politicized market. If Nvidia goes neutral, neither lab has a hardware patron. Watch who fills that vacuum.
Meanwhile, the Gemini wrongful death lawsuit and the AI relicensing debate are two different faces of the same problem: we have no legal framework for AI's consequences. A man is dead, and the question of whether a chatbot bears legal responsibility remains genuinely unsettled. Code can be "relicensed" by running it through an AI rewriter, and nobody knows if that's legal. The law is years behind.
The Qwen exodus continues — three senior departures this year. If China's most promising open-source AI lab is hemorrhaging talent due to internal politics, that's a gift to Western frontier labs and a headwind for the open-source movement's strongest non-Western contributor.
Watch today: Whether OpenAI responds publicly to Amodei's "lies" accusation. If Altman stays quiet, it's because the Pentagon deal is too important to jeopardize with a PR war. If he responds, the gloves are truly off.