Anthropic Accidentally Releases Claude Code Source Files
The slip-up sent developers hunting for clues and sharpened concerns about security and disclosure practices
Topics
News
- India’s Workers Are Rushing to Reskill as AI Reshapes Jobs
- Anthropic Accidentally Releases Claude Code Source Files
- AI Becomes Boardroom Priority for Family Businesses, Deloitte Report Finds
- IndiaAI Chief Abhishek Singh Moved to Head Testing Agency
- Oracle Begins Deep Job Cuts Amid AI Spending Crunch
- Google Rolls Out Cheaper Video AI Model Veo 3.1 Lite
Anthropic accidentally released internal source code for Claude Code, its AI coding tool, Bloomberg reported, in the latest disclosure mishap at the startup behind the Claude chatbot.
The release sent developers through the files for clues about how the product works and what the company may build next, according to Bloomberg.
Anthropic’s release quickly set off a burst of discussion online, with developers combing through the files for clues about unreleased features, memory systems and how Claude Code handles tasks.
Others raised security concerns, warning that even if no customer data or model weights were exposed, the code could still offer rivals and researchers a closer look at how the product is built.
“Earlier today, a Claude Code release included some internal source code. No sensitive customer data or credentials were involved or exposed,” an Anthropic spokesperson told Bloomberg in a statement. “This was a release packaging issue caused by human error, not a security breach. We’re rolling out measures to prevent this from happening again.”
The release raised fresh questions about Anthropic’s internal controls, with developers and security researchers warning that even a limited code exposure can reveal how a product is built and where weaknesses may lie.
Bloomberg said the exposed material related to Claude Code rather than Anthropic’s underlying model weights or customer systems.
The mistake came less than a week after another Anthropic disclosure problem.
Fortune reported that thousands of files linked to the company had been left publicly accessible, including a draft blog post describing an upcoming model known internally as “Mythos” and “Capybara.”
