Anthropic’s Fable 5 was restored after a 19-day government-ordered suspension, the same day OpenAI proposed ceding 5% equity to a US sovereign wealth fund and Microsoft unveiled a $2.5B enterprise AI deployment unit. The overarching theme: frontier AI labs are actively courting — and entangling themselves with — both Washington and the Fortune 500.
The Trump administration lifted export controls on Anthropic’s Fable 5 and Mythos 5 models, ending a weeks-long standoff and restoring global access starting today. Cloudflare gave AI companies a September 15 deadline to separate training crawlers from search crawlers or face default blocks on publisher sites. On the science front, Realta Fusion demonstrated the first direct electricity generation from a nuclear fusion reaction, while Conception produced the world’s first human eggs from stem cells.
Anthropic dominates the day with two major launches — Claude Sonnet 5, a 1M-context agentic model at steep promotional pricing, and Claude Science, a unified research workbench for scientists. Meanwhile, AI chip startup Etched emerges from stealth with $800M raised and $1B in customer contracts, and the Amazon–Anthropic relationship grows complicated.
OpenAI previewed its GPT-5.6 family — Sol, Terra, and Luna — in a government-supervised limited release to roughly 20 approved partners. Meanwhile, Anthropic finds itself simultaneously banned from federal use and celebrated by California, and China’s GLM-5.2 claims cybersecurity parity with frontier models.
A Palisades fire arson trial ended in mistrial after prosecutors made history using ChatGPT conversation logs as evidence — a legal first. Anthropic escalated its geopolitical fight against AI theft, accusing Alibaba-linked entities of the largest known distillation attack. Meanwhile, Wall Street crowned Micron the next Nvidia as the AI memory supercycle accelerated.
The Trump administration granted ~100 US companies and agencies access to Anthropic’s Mythos 5, ending a two-week standoff — but Fable 5 remains off-limits to foreign users. Meanwhile, Asian startups are racing to fill the void, OpenAI previews GPT-5.6, and Google’s Gemini 2.5 Pro resets the reasoning leaderboard.
OpenAI unveiled GPT-5.6 in a restricted preview after the Trump administration intervened, while Anthropic’s Mythos-class models remain offline for a second week with no resolution in sight. Chinese rival Z.ai’s GLM-5.2 is quietly filling the vacuum.
OpenAI and Broadcom unveiled Jalapeño, OpenAI’s first custom inference chip, marking a decisive step toward full-stack AI ownership. Google extended computer-use capabilities to Gemini 3.5 Flash, while new data shows Claude steadily pulling paid subscribers away from ChatGPT.
OpenAI unveiled Jalapeño, its first in-house AI inference chip built with Broadcom, on a day dominated by revelations that Anthropic’s Mythos model reportedly breached nearly all NSA classified systems within hours — the incident behind last week’s landmark AI export controls. Agility Robotics also announced a $2.5B SPAC deal to become the first publicly traded pure-play humanoid company.
Anthropic launched Claude Tag, making Claude a persistent AI teammate inside Slack with shared organizational memory and ambient proactivity. OpenAI escalated its cybersecurity push with GPT-5.5-Cyber and the Patch the Planet open-source initiative. SpaceX signed a $6.3B compute deal with Reflection AI, deepening its role as an AI infrastructure provider.