A nuclear-risk classifier now runs on live Claude traffic at 96% preliminary accuracy, adding an additional filter for nuclear-adjacent workloads. The co-development method is being published as a template for other labs.

• articleAnthropic Invests 50 Billion In American Ai Infrastructure
• articleDeveloping Computer Use
• articleClaude In Xcode
• articleUpdates To Our Consumer Terms

• article[AINews] "Laguna S 2.1 Released: Cheaper than Deepseek v4 Flash, Better than V4 Pro"Latent Space
• articleStrategic Warning For Ai Risk Progress And Insights From Our Frontier Red TeamAnthropic News
• podcast8/6: Cyber Risk and Bio RiskTheo Jaffee

Checking sign-in…

Loading comments…