NEARLY 12K API KEYS AND PASSWORDS FOUND IN AI TRAINING DATA

Summary from the TLDR InfoSec newsletter

Researchers at Truffle Security scanned 400TB of data from the Common Crawl December 2024 archive. Common Crawl is a non-profit that regularly publishes massive crawled web data datasets frequently used to train LLMs. The researchers identified 219 distinct secret types, including valid AWS, MailChimp, and WalkScore API keys.