Anthropic's Mythos Is Too Dangerous to Release, Revenue Tripled, Open Source Hits the Frontier

Today we're covering the biggest AI stories of April 8th, 2026. Anthropic unveiled Project Glasswing — a defensive cybersecurity coalition with Apple, Google, Microsoft, AWS, and Nvidia built around Claude Mythos Preview, a model so powerful the company decided not to release it publicly, instead giving access to select partners to patch critical security flaws before similar capabilities fall into the wrong hands. Despite facing a Pentagon blacklist, Anthropic tripled its run-rate revenue to $30 billion since January and doubled its million-dollar enterprise customers to over 1,000 — locking in a 3.5 gigawatt compute deal with Google and Broadcom to keep up with demand. Plus, Chinese open-source lab Z AI released GLM-5.1, which just hit number one on a top coding benchmark above both GPT-5.4 and Opus 4.6 — and a community workflow from a pickleball student who built a custom AI coach to practice strategy, scoring, and rules between lessons.