Anthropic says it accidentally leaked source code for Claude Code, compounding earlier reported disclosure issues involving an internal upcoming model. The new exposure included hundreds of thousands of lines across roughly 1,900 files, according to reporting reviewed by cybersecurity specialists. Anthropic attributed the problem to a packaging release issue tied to “human error,” adding that no sensitive customer data or credentials were exposed. The leaked agentic harness source code could still enable competitors to reverse-engineer elements of how Claude Code orchestrates tools. Higher education technology leaders should treat the incident as a reminder to verify release engineering, access controls, and third-party AI governance—especially as universities increasingly rely on agentic coding and automation tools in research and academic operations.