Show HN: On-device Chrome extension that blocks credential leaks to LLM chats (redact.clearformlabs.com)

A new Chrome extension called Redact has been launched, aimed at preventing the accidental sharing of sensitive information such as credentials and personally identifiable information (PII) during interactions with large language models (LLMs) like ChatGPT and Claude. By intercepting paste actions in real-time, Redact runs a local neural network in the user's browser to swiftly scan and identify sensitive data before it can be sent. Users are given the option to block, warn, or send a redacted version of the paste, ensuring that sensitive information remains protected. This development is significant for the AI/ML community as it addresses a critical vulnerability in using LLMs—a growing concern over data privacy and security. The local detection process, which completes in about 150 milliseconds, eliminates the need for external server interactions, thus preserving user privacy while maintaining transparency; the extension is fully auditable with its source code available on GitHub. Designed to work with major chat platforms, Redact empowers users to safeguard their data easily and efficiently, representing a proactive step toward enhancing security in AI interactions.