A security firm, Apiiro, discovered that Anthropic's Claude Code assistant inadvertently saved sensitive credentials, including passwords and API keys, in a configuration file. Developers who clicked the "don't ask again" option when prompted to approve shell commands allowed these secrets to be permanently stored. Apiiro found over 100,000 copies of this configuration file on public GitHub, with over 1,100 containing actual secrets, potentially exposing data for numerous organizations. AI
IMPACT Highlights the critical need for robust security practices and secret management when using AI coding assistants, as convenience features can inadvertently expose sensitive data.
RANK_REASON Security vulnerability discovered in an AI coding assistant's functionality.
Read on dev.to — Claude Code tag →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →