AI company Anthropic PBC recently accidentally released the internal source code of its popular coding assistant Claude, triggering widespread industry concern over its security practices. The company, which brands itself on prioritizing security, stated the incident originated from a human error caused by a release packaging issue, involving approximately 1,900 files and 512,000 lines of code from the Claude Code tool. Anthropic emphasized this was not a security vulnerability, but it represents the second security lapse in just a few days, damaging its reputation.
The code leak was first exposed on social media platform X, garnering over 30 million views, and included key systems like Claude Code's context management pipeline. AI cybersecurity firm Straiker warned that attackers could analyze data flows from the leak, with potential risks including backdoor implantation and long-term session attacks. Although Anthropic executives blamed a "process error," the scale of the disclosure has led developers to worry about its operational security and intellectual property protection.
