The Download: biotech’s future and cheaper, cleaner steel

Details

This is a newsletter digest from MIT Technology Review covering several stories from September 11, 2026. The most significant AI-related disclosure: Anthropic identified and blocked five cases where users attempted to use Claude to assist in biological weapon development, and six additional cases involving conventional weapons software creation. The company also disclosed that Russia-linked hackers used Claude to automate cyberattack operations targeting Ukraine, and that some governments are actively deploying Claude for surveillance purposes. These disclosures come from a new Anthropic safety report — the article links to coverage in the NYT, BBC, Axios, Quartz, and The Guardian, but provides limited detail itself. The digest also spotlights MIT Technology Review's 35 Innovators Under 35 list for 2026, featuring nine biotech honorees. Their work includes a cellular reprogramming therapy that reverses vision loss, neural electrodes inspired by Japanese paperfolding art, a personalized gene-editing treatment developed for a specific infant with a rare genetic disorder, and generative AI systems designed to produce novel therapeutic viruses. One honoree, Laureen Meroueh of Hertha Metals, is developing a single-step steelmaking furnace that replaces coal with natural gas, claiming at least 50% emissions reduction and 25% cost savings. As a newsletter roundup, depth on any individual story is limited — the piece is best read as a pointer to primary sources. The Anthropic safety disclosures are the most consequential AI-relevant items in this edition.