Anthropic appeared before a federal judge in San Francisco to fight a White House designation labeling it a "national security supply chain risk" — imposed after the company refused to allow Claude to be used for lethal autonomous weapons without human oversight. OpenAI, Google, and Microsoft all filed supporting briefs for Anthropic. This is the first major courtroom confrontation between an AI lab and the federal government over weapons use.
The administration released a seven-pillar framework pushing Congress toward unified federal AI legislation — and explicitly calling to preempt all state-level AI laws. Rather than a new regulatory agency, the framework leans on existing sector-specific regulators, giving tech companies the looser oversight regime they've lobbied for.
Meta's chief AI scientist Yann LeCun launched Advanced Machine Intelligence Labs with the largest seed round in European venture history, backed by Nvidia, Bezos, Samsung, and Temasek. The company is betting on "world models" via JEPA architecture — LeCun's long-running thesis that transformers and LLMs are a dead end for reaching genuine AI understanding.
Alibaba released Qwen 3.5 with multimodal support including video up to two hours long, joining simultaneous upgrades from Baidu, ByteDance, Tencent, and MiniMax. MiniMax's M2.5 is drawing particular attention for reportedly matching Claude Opus 4.6 at a fraction of the cost — another sign that the performance gap between Chinese and Western frontier models is closing fast.
SCOTUS declined to hear Thaler v. Perlmutter, letting stand the ruling that AI-generated works without human authorship cannot be copyrighted. The decision closes a major legal door for developers hoping to monetize purely machine-generated content and cements the Copyright Office's human-authorship requirement as the law of the land.
The two reference publishers filed suit alleging OpenAI scraped nearly 100,000 of their copyrighted articles to train ChatGPT without authorization. They also argue that ChatGPT fabricates content users mistakenly attribute to their brands — adding a reputational harm claim on top of the training data claim.
Broadcom disclosed that TSMC's manufacturing capacity — not chip design or software — has become the primary constraint on AI hardware supply. With demand for custom AI accelerators still surging, this signals that the next major scaling wall isn't algorithmic; it's physical.
PauseAI and allied groups organized a San Francisco march calling for a coordinated halt to frontier AI development, stopping at all three major lab offices. Organizers proposed international treaties and compute caps as verification mechanisms — niche proposals today, but a preview of the political friction ahead as models grow more capable.
