Anthropic Accidentally Leaks 512,000 Lines of Claude Code Source Code

Key Points

• Approximately 1,900 files and 512,000 lines of Claude Code source code were accidentally released
• This is Anthropic's second security incident in days
• Security researchers warn attackers could craft persistent backdoors by studying the four-stage context management pipeline
• The post sharing the code link garnered over 30 million views on X

Full Details

Anthropic PBC, the AI safety-focused company behind the Claude assistant, accidentally released internal source code for its Claude Code coding tool. The release compromised approximately 1,900 files and 512,000 lines of code related to Claude Code, an agentic coding tool that runs directly inside developer environments. The incident first came to light on social media platform X, where a post sharing a link to the code garnered over 30 million views. Security researchers from Straiker warned that attackers can now study and fuzz exactly how data flows through Claude Code's four-stage context management pipeline and craft payloads designed to survive compaction, effectively persisting a backdoor across arbitrarily long sessions. This is Anthropic's second security slip-up in a matter of days, raising questions about the security practices of an AI company that has built its brand on prioritizing safety.