Descriptions:
Wes Roth covers what was, at time of recording, a live 48-hour countdown to a Pentagon-imposed deadline for Anthropic to grant unrestricted military access to Claude or face designation as a supply chain risk. The video traces how the crisis originated: leaked reports revealed that Claude was used through Palantir’s Maven Smart System during a covert Joint Special Operations Command (JSOC) raid in Caracas, Venezuela, which resulted in the capture of former president Nicolás Maduro. This use case — lethal operational support — was precisely the scenario Anthropic’s safety policies were designed to prevent.
A central revelation in the video is Anthropic’s quiet publication of RSP 3.0, its updated Responsible Scaling Policy, which removes a core commitment the company had held since 2023: a categorical pledge never to train a model without verified safety guarantees in place. Chief Science Officer Jared Kaplan explained the change to Time magazine, framing it as a pragmatic adjustment to competitive reality. Roth analyzes what the softened dual-condition replacement signals about Anthropic’s shifting posture under government and competitive pressure.
The video also outlines the full scope of consequences a supply chain risk designation would carry: likely cancellation of Anthropic’s planned IPO, forced removal from defense contractor workflows at companies like Lockheed Martin and Boeing, and broader reputational damage. Perspectives from Pentagon CTO emails, retired General Jack Shanahan, and Anthropic’s own statements are used to frame what Roth characterizes as a historic confrontation between a leading AI safety organization and the US defense establishment.
📺 Source: Wes Roth · Published February 26, 2026
🏷️ Format: News Analysis
