Anthropic AI Updates: April 1, 2026

1. Claude Code Source Code Exposed via npm Packaging Error

Anthropic accidentally published the full source code of Claude Code to the public npm registry when version 2.1.88 shipped with a 59.8 MB JavaScript source map file intended for internal debugging. Security researcher Chaofan Shou discovered the leak, which exposed approximately 512,000 lines of TypeScript across roughly 1,900 files. The exposed code revealed internal codenames for unreleased models, anti-distillation mechanisms, an “undercover mode” that strips Anthropic branding in external repos, and a bug causing an estimated 250,000 wasted API calls daily. Anthropic called it “a release packaging issue caused by human error, not a security breach” and confirmed no customer data or model weights were compromised. This marked Anthropic’s second accidental data exposure in one week, following the Mythos model spec leak on March 26. Source

2. Anthropic Signs AI Safety and Economic Data Sharing Deal with Australia

Anthropic agreed to share its economic index data with the Australian government to help track AI adoption across the economy and its impact on workers and jobs. Under the deal, Anthropic will share findings on emerging AI model capabilities and risks, participate in joint safety evaluations, and collaborate on research with Australian universities. CEO Dario Amodei traveled to Canberra and was expected to meet Prime Minister Anthony Albanese to finalize the agreement. Source