Summary
Nate Herk runs through the defining AI news of the week ending September 13th, 2026, starting with Jacob Coxon’s resignation from Anthropic — a tweet that hit 150 million views within days. The video goes deeper than most coverage by highlighting Evan Hubinger, Anthropic’s alignment science lead, who quote-tweeted Coxon with a stark statement: he personally believes there is greater than a 10% chance AI kills all humans within the next decade, and that Anthropic does not yet have a plan to solve superintelligence alignment.
Nate contextualizes Dario Amodei’s subsequent 9,000-word essay as a response to this public pressure, focusing on the “race to the bottom” incentive problem — every lab is pushed to ship faster than its competitors, often at the expense of safety. Amodei’s proposed solution is to make safety itself a competitive axis, creating a “race to the top.”
The video also covers Anthropic’s newly released AI misuse report, which documents harms across seven categories including cyber operations and AI-powered scams. Two allegations stand out: Anthropic attributes 151 million unauthorized exchanges to Alibaba’s model training campaign, and accuses Moonshot and DeepSeek of secretly routing traffic through Claude to extract outputs. The second half situates all of this against continuous Chinese model releases from DeepSeek, Qwen, GLM, and Minimax, asking whether the safety coalition is as much about competitive anxiety as genuine alignment concern.
📺 Source: Nate Herk | AI Automation · Published September 13, 2026
🏷️ Format: News Analysis
