Anthropic accidentally exposed part of Claude Code's internal source code

In a surprising turn of events, Anthropic has unintentionally revealed part of the internal source code for its highly-regarded AI coding assistant, Claude Code. This incident came to light following a release, confirmed by a spokesperson to Business Insider on Tuesday. The spokesperson clarified, "Earlier today, a Claude Code release included some internal source code. No sensitive customer data or credentials were involved or exposed." They emphasized that this leak resulted from a packaging error caused by human oversight, rather than a security breach. To mitigate future risks, the company is implementing new measures. The leaked code, which has drawn significant attention, was shared in a post on X, garnering an impressive 26 million views by Tuesday evening. It pertains specifically to Claude Code and does not include the foundational AI models. This exposure may provide Anthropic's competitors with valuable insights into one of its flagship products, potentially giving them a competitive edge. Furthermore, this incident raises important questions about the security protocols at a company that prides itself on prioritizing AI safety. The leak follows a period of notable growth for Anthropic after a public dispute with the Pentagon earlier this year. CEO Dario Amodei stood firm in negotiations regarding the usage of their AI technology, resulting in the Defense Department opting to partner with OpenAI instead. In the wake of this controversy, Anthropic's Claude chatbot experienced a significant spike in downloads, briefly reaching the top position in the US Apple App Store last week, just as a US District Judge granted a temporary injunction that impacted supply chain risk designations.