Within days of each other, Anthropic first leaked the source code to Claude Code, and then a critical vulnerability was found by Adversa AI. On March 31, 2026, Anthropic mistakenly included a ...